What do you think?
Rate this book
AI - Bước Tiến Đột Phá Trong Chăm Sóc Sức Khỏe
Eric Topol là một trong những bác sỹ giỏi nhất nước Mỹ, chuyên về lĩnh vực tim mạch, đồng thời là người đứng đầu Viện Khoa học Tịnh tiến Scripps, nguyên là Chủ tịch tim mạch của Cleveland Clinic. Ông được liệt vào top 10 nhà khoa học có những bài báo được trích dẫn nhiều nhất hiện nay.
Nền Y học trong thời đại chúng ta đang sống đã chạm ngưỡng báo động về sự thiếu liên kết giữa con người với nhau, cụ thể là bác sĩ với bệnh nhân.
AI xuất hiện cùng với những hứa hẹn đầy triển vọng về một tương lai của ngành Y, nơi bác sĩ giao tiếp, đồng cảm, thấu hiểu với bệnh nhân bằng tất cả lòng trắc ẩn của mình.
AI, một mặt, trả lại tính nhân văn sâu sắc cho Y học, đưa Y học trở thành “Deep medicine” nếu được sử dụng đúng cách ; mặt khác, nếu sử dụng sai cách thì AI sẽ dẫn Y học của chúng ta tới vực thẳm.
Làm thế nào để AI đưa Y học phát triển theo đúng lộ trình, sự kết nối giữa bác sỹ và bệnh nhân ngày một thắt chặt... Câu trả lời sẽ được bác sĩ hàng đầu Hoa Kỳ Eric Topol giải đáp trong Deep Medicine: AI – Bước tiến đột phá trong chăm sóc sức khỏe.
Lời bình
“Y học là nơi tạo ra cơ hội tốt nhất cho sự kết hợp cộng sinh giữa trí tuệ nhân tạo và con người. AI – Bước tiến đột phá trong chăm sóc sức khỏe của Eric Topol đưa ra những quan điểm mạnh mẽ từ một người hiểu sâu sắc cả lĩnh vực chăm sóc sức khỏe và AI." Eric Topol, Tác giả
"Tôi rất khuyến khích cuốn sách và hy vọng nó kết nối các bác sĩ y khoa và các nhà nghiên cứu AI, và giúp họ hiểu rằng chỉ khi làm việc cùng nhau, chúng ta mới có thể đạt được ước mơ về sức khỏe và tuổi thọ con người.” - Kai-Fu Lee, nhà đầu tư AI được biết đến nhiều nhất hành tinh
Nền Y học trong thời đại chúng ta đang sống đã chạm ngưỡng báo động về sự thiếu liên kết giữa con người với nhau, cụ thể là bác sĩ với bệnh nhân.
AI xuất hiện cùng với những hứa hẹn đầy triển vọng về một tương lai của ngành Y, nơi bác sĩ giao tiếp, đồng cảm, thấu hiểu với bệnh nhân bằng tất cả lòng trắc ẩn của mình.
AI, một mặt, trả lại tính nhân văn sâu sắc cho Y học, đưa Y học trở thành “Deep medicine” nếu được sử dụng đúng cách ; mặt khác, nếu sử dụng sai cách thì AI sẽ dẫn Y học của chúng ta tới vực thẳm.
Làm thế nào để AI đưa Y học phát triển theo đúng lộ trình, sự kết nối giữa bác sỹ và bệnh nhân ngày một thắt chặt... Câu trả lời sẽ được bác sĩ hàng đầu Hoa Kỳ Eric Topol giải đáp trong Deep Medicine: AI – Bước tiến đột phá trong chăm sóc sức khỏe.
Lời bình
“Y học là nơi tạo ra cơ hội tốt nhất cho sự kết hợp cộng sinh giữa trí tuệ nhân tạo và con người. AI – Bước tiến đột phá trong chăm sóc sức khỏe của Eric Topol đưa ra những quan điểm mạnh mẽ từ một người hiểu sâu sắc cả lĩnh vực chăm sóc sức khỏe và AI." Eric Topol, Tác giả
"Tôi rất khuyến khích cuốn sách và hy vọng nó kết nối các bác sĩ y khoa và các nhà nghiên cứu AI, và giúp họ hiểu rằng chỉ khi làm việc cùng nhau, chúng ta mới có thể đạt được ước mơ về sức khỏe và tuổi thọ con người.” - Kai-Fu Lee, nhà đầu tư AI được biết đến nhiều nhất hành tinh
482 pages, Paperback
First published January 1, 2019
1141 people are currently reading
7969 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 285 reviews
January 21, 2021
As an MD who now works on Healthcare IT field, I should start by telling that I am a big fan of Eric Topol and admire his vision in many ways. I read his previous books, follow him on twitter and watched him live multiple times in various events. When I was a TED translator, I translated couple of his TED talks to Turkish ( and he even acknowledged it in a signed book when I met him). So I started to read this with a huge positive bias obviously.
I liked the book overall. He is giving a nice summary of the current issues in healthcare, the current status of AI, the update on the AI research related to healthcare, its various use in different areas, and finally provides a summary of where it might take us.
Things disappointed me was to see the same cliche reasoning around the EMRs. Yes, physician burn out is a big problem, and we should be worrying about it, and should try to find solutions to remediate it. But like many others Topol puts all the blame to EMRs, and repeats the same mantra of "how EMRs are the cause of all burnout" and how "they are designed for billing not clinical care". Now as a person who works in this field I can attest that the latter is not true, at least not anymore. The billing focused EMR is a thing of the past - like 20 years or so, and those days are long past but it is an easy target so everyone use the same mantra rather than acknowledging the complex issues around healthcare, its finance, lack of appropriate and wide coverage primary & preventive care, the policies, rules, regulations and finally IT.
The first claim - EMRs being the cause of all burnout is also dubious in my opinion. I am sure they do have an impact and definitely it is a factor that impacts patient/provider interaction. However, there are other causes like regulations, the monetary healthcare system, fee for service and faulty incentives. ( As an anecdote, in my first EMR implementation job in the US, I was shocked to find out that the private practice I was working with was scheduling doctors so they see two patients at the same 15 min interval. When they were doing this they were on paper, and had no EMR at all. I could not even wrap my brain around how a doctor can see two patients in the same 15 minutes and advised against it. So no, it was not the EMRs who disconnected the patients from doctors, it was the healthcare system itself. And please do not tell me that providers were able to prep and chart properly for those patients they saw on average of 7.5 minutes each when they were on paper. They weren't and the only difference is that nobody knew. Now it is all in the open.)
I would anticipate a more balanced approach from Topol about EMRs, for example pointing out couple of benefits of EMRs in addition to the challenges they provided and was disappointed to read his bias over and over.
The second drawback for me was the endorsement of couple new healthcare tech companies with dubious claims. He did the same thing in his previous book "The patient will see you now" about Theranos. He told how amazing this technology and how impressed he was for pages. And we all know how Theranos ended up...
He does the same thing for two new companies DayTwo and Viome, and in a similar way promotes a pretty ambitions claim that is yet to be proven and accepted. Both companies are in search of more investors and the scientist are vary of wild claims made by these kinds of startups. So it was a real turn off for me when I read an almost advertorial promotion of those, similar to what he did for Theranos.
These two issues made me deduce two stars since I felt there was a certain bias in some of the issues and how they are being narrated.
Other than these two, I think it is a good compilation of AI and its use in the industry and it's potentials. I especially liked the section on AI Applications on Mental Health.
Definitely a must read for those who are interested in AI, Medicine and Health Informatics. But read it with a grain of salt from time to time.
I liked the book overall. He is giving a nice summary of the current issues in healthcare, the current status of AI, the update on the AI research related to healthcare, its various use in different areas, and finally provides a summary of where it might take us.
Things disappointed me was to see the same cliche reasoning around the EMRs. Yes, physician burn out is a big problem, and we should be worrying about it, and should try to find solutions to remediate it. But like many others Topol puts all the blame to EMRs, and repeats the same mantra of "how EMRs are the cause of all burnout" and how "they are designed for billing not clinical care". Now as a person who works in this field I can attest that the latter is not true, at least not anymore. The billing focused EMR is a thing of the past - like 20 years or so, and those days are long past but it is an easy target so everyone use the same mantra rather than acknowledging the complex issues around healthcare, its finance, lack of appropriate and wide coverage primary & preventive care, the policies, rules, regulations and finally IT.
The first claim - EMRs being the cause of all burnout is also dubious in my opinion. I am sure they do have an impact and definitely it is a factor that impacts patient/provider interaction. However, there are other causes like regulations, the monetary healthcare system, fee for service and faulty incentives. ( As an anecdote, in my first EMR implementation job in the US, I was shocked to find out that the private practice I was working with was scheduling doctors so they see two patients at the same 15 min interval. When they were doing this they were on paper, and had no EMR at all. I could not even wrap my brain around how a doctor can see two patients in the same 15 minutes and advised against it. So no, it was not the EMRs who disconnected the patients from doctors, it was the healthcare system itself. And please do not tell me that providers were able to prep and chart properly for those patients they saw on average of 7.5 minutes each when they were on paper. They weren't and the only difference is that nobody knew. Now it is all in the open.)
I would anticipate a more balanced approach from Topol about EMRs, for example pointing out couple of benefits of EMRs in addition to the challenges they provided and was disappointed to read his bias over and over.
The second drawback for me was the endorsement of couple new healthcare tech companies with dubious claims. He did the same thing in his previous book "The patient will see you now" about Theranos. He told how amazing this technology and how impressed he was for pages. And we all know how Theranos ended up...
He does the same thing for two new companies DayTwo and Viome, and in a similar way promotes a pretty ambitions claim that is yet to be proven and accepted. Both companies are in search of more investors and the scientist are vary of wild claims made by these kinds of startups. So it was a real turn off for me when I read an almost advertorial promotion of those, similar to what he did for Theranos.
These two issues made me deduce two stars since I felt there was a certain bias in some of the issues and how they are being narrated.
Other than these two, I think it is a good compilation of AI and its use in the industry and it's potentials. I especially liked the section on AI Applications on Mental Health.
Definitely a must read for those who are interested in AI, Medicine and Health Informatics. But read it with a grain of salt from time to time.
May 11, 2019
Critics hailed Big Data: A Revolution That Will Transform How We Live, Work, and Think as a vision of future for big data. But a lot of it turned out to be hype. I see the same hubris in health care AI. Take IBM Watson for example, IBM Watson still hasn't delivered much in health care. (https://spectrum.ieee.org/biomedical/...)
We have seen the mess created by "move fast and break things" philosophy. The way I see it, Deep medicine might be the "move fast and break things" in health care.
Certainly Artificial Intelligence will provide a helping hand in health care. Take diabetic retinopathy, the number one global cause of vision loss. If all the recommended screening of people with diabetes was performed, there would be well over 30 million retinal images per year that would need to be evaluated. Clearly, this sounds like a job for deep learning.
I was mesmerized when I read this line: "with the Aipoly app, a senior with significant visual impairment can simply point to an object with a smartphone and AI will quickly kick in with a voice response identification." But on second thought I am not that impressed because it's not a cure — it's just "if all you have is a hammer, everything looks like a nail" — most of the anecdotes mentioned in this book are based on mobile applications, because most of the software engineers see the smartphone as the sole solution to every problem. Since most startups are aiming to use smartphones as a diagnostic tool, it might result in "overdiagnosis". Artificial intelligence cannot replace actual science, it can help speed things up, it is a complimentary tool but it can never be a replacement for actual science.
It reminded me of an anecdote told of the great American inventor Thomas Edison. A practical man with little formal education, Edison nevertheless understood the value of education but also never missed a chance to show how a clever person could often work around a technical deficiency. For example, after hiring a young mathematician Edison assigned him the task of determining the volume of a new lightbulb, a bulb designed with an undulating shape. The mathematician carefully reduced the shape to a complicated equation and then laboriously, over a period of hours, integrated the equation over three dimensions to get the volume enclosed. Then, he proudly showed the result to Edison.
Edison congratulated the man on being a fine mathematician, as his computed answer agreed quite well with Edison’s own value, which he had arrived at in less than 30 seconds. When the astonished mathematician asked how Edison had done that, the inventor (without saying a word) simply filled the bulb with water and then poured the water out of the bulb into a glass beaker with volume levels marked on the side.
Edison had made his point: math is great, but use it as a tool and not as a crutch. Same with artificial intelligence: we should definitely use it as a tool but not as a crutch.
"A gold mine is only as good as the man digging." Artificial intelligence depends on good unbiased data. Like facial recognition software that detects only white faces because the data it was trained on did not contain black faces, Deep medicine might inflate bias that already exists in the health care.
And finally like all things pertained with the information age there is the issue of privacy.
We have seen the mess created by "move fast and break things" philosophy. The way I see it, Deep medicine might be the "move fast and break things" in health care.
Certainly Artificial Intelligence will provide a helping hand in health care. Take diabetic retinopathy, the number one global cause of vision loss. If all the recommended screening of people with diabetes was performed, there would be well over 30 million retinal images per year that would need to be evaluated. Clearly, this sounds like a job for deep learning.
I was mesmerized when I read this line: "with the Aipoly app, a senior with significant visual impairment can simply point to an object with a smartphone and AI will quickly kick in with a voice response identification." But on second thought I am not that impressed because it's not a cure — it's just "if all you have is a hammer, everything looks like a nail" — most of the anecdotes mentioned in this book are based on mobile applications, because most of the software engineers see the smartphone as the sole solution to every problem. Since most startups are aiming to use smartphones as a diagnostic tool, it might result in "overdiagnosis". Artificial intelligence cannot replace actual science, it can help speed things up, it is a complimentary tool but it can never be a replacement for actual science.
It reminded me of an anecdote told of the great American inventor Thomas Edison. A practical man with little formal education, Edison nevertheless understood the value of education but also never missed a chance to show how a clever person could often work around a technical deficiency. For example, after hiring a young mathematician Edison assigned him the task of determining the volume of a new lightbulb, a bulb designed with an undulating shape. The mathematician carefully reduced the shape to a complicated equation and then laboriously, over a period of hours, integrated the equation over three dimensions to get the volume enclosed. Then, he proudly showed the result to Edison.
Edison congratulated the man on being a fine mathematician, as his computed answer agreed quite well with Edison’s own value, which he had arrived at in less than 30 seconds. When the astonished mathematician asked how Edison had done that, the inventor (without saying a word) simply filled the bulb with water and then poured the water out of the bulb into a glass beaker with volume levels marked on the side.
Edison had made his point: math is great, but use it as a tool and not as a crutch. Same with artificial intelligence: we should definitely use it as a tool but not as a crutch.
"A gold mine is only as good as the man digging." Artificial intelligence depends on good unbiased data. Like facial recognition software that detects only white faces because the data it was trained on did not contain black faces, Deep medicine might inflate bias that already exists in the health care.
And finally like all things pertained with the information age there is the issue of privacy.
October 6, 2022
Though this is a AI book---algorithms---just because AI can drive a car better than you is not a reason why it should take over driving. It's not about the competition of the human vs. the robot, yet more of a collaboration between the two. There is a reason of course to see your physician (though this is becoming increasingly difficult) given the burden on healthcare practitioners today. Face to face or Telehealth visits are a substantial part of wellness and care. Dr. Eric J. Topol (Cardiologist) who states the average physician spends 7 minutes per patient. Topols' outlook is to make healthcare human again.
"AI can revolutionize other aspects of our lives that are, in one sense or another, upstream from the clinic. A huge one is how we eat. One of the unexpected and practical accomplishments of machine learning to date has been to provide a potential scientific basis for individualized diets."
---Dr. Eric J. Topol
Though I cannot attest to a 7 minute visit, I can say in certain areas this is all that can be afforded by the overwhelmed physician in a place where demographics prevent them from administering proper care. The Doctor patient relationship is sacrosanct to health as "It is more important to know what sort of person has a disease than to know what sort of disease a person has---Hippocrates." Some characteristics of the human condition that AI cannot accomplish are, empathy, human touch, love, laughing, crying, telling stories, jokes, generosity and intuitiveness to the needs of a human being.
"AI can revolutionize other aspects of our lives that are, in one sense or another, upstream from the clinic. A huge one is how we eat. One of the unexpected and practical accomplishments of machine learning to date has been to provide a potential scientific basis for individualized diets."
---Dr. Eric J. Topol
Though I cannot attest to a 7 minute visit, I can say in certain areas this is all that can be afforded by the overwhelmed physician in a place where demographics prevent them from administering proper care. The Doctor patient relationship is sacrosanct to health as "It is more important to know what sort of person has a disease than to know what sort of disease a person has---Hippocrates." Some characteristics of the human condition that AI cannot accomplish are, empathy, human touch, love, laughing, crying, telling stories, jokes, generosity and intuitiveness to the needs of a human being.
This entire review has been hidden because of spoilers.
January 8, 2020
Kas tehisintellekt (AI), mille võimet muuta ja kujundada tuleviku on võrreldud elektri ja auruveduriga, muudab tervishoiusüsteemi? Absoluutselt. Kas AI-l on võimalus vähendada tervishoiukulutusi? Absoluutselt. Arvestades isegi asjaoluga, et rahvastik vananeb, eeldused tervishoiukvaliteedi osas suurenevad ning uute ravimite maksumus hakkab al 100 000$? Jah.
Ühiskonnana oleme leiutanud väga võimsa tööriista, mis on võimeline leidma tohututest andmemahtudest seaduspärasusi, milleks ükski inimene võimeline pole. Alustan lihtsamatest näidesest ja hakkan suurendama vajaminevaid andmemahtusid.
Aktiivsusmonitor detekteerib südame löögisageduse tõusu ilma, et inimese füüsiline aktiivsus oleks tõusnud. Palub koheselt teha inimesel EKG (elektrokardiogramm), detekteerib südame rütmihäire. EKG läheb arstile, patsient saab varakult arsti juurde. Arst määrab vajadusel ravimid.
Toidame algoritmi kümneid tuhandeid EKG-d, AI leiab seaduspärasusi, mille alusel on võimalus mitte-invasiivselt määrata inimese kaaliumi taset. Eluliselt oluline näitaja neerukahjustusega patsientideke.
Toidame närvivõrgustikku inimese DNA ning leiame seoseid geenide ja haiguste (nt autismi, diabeedi, Alzheimeri jt haiguste vahel) vahel.
Suurendame andmemahtusid veelgi. Toidame algoritmi andmeid, mis puudutavad: inimese sotsiaalseid ja käitumuslike mustreid, genoomikat jt 'oomikaid' (transkriptoomika, metaboloomika, proteoomika jt), immuunsüsteemi, mikrobioomi, anatoomi, keskkonna info, terviselugu, inimese kõne, biosensorite jt andmed, mis võimaldab meil saada väga hea ülevaate inimese tervise seisundist. Arvestades seda ravida haigust personaalsel tasemel. Kuid peamine, mis me selle andmestikuga teha saaksime, on muuta inimese käitumist, et vältida haiguste teket.
Kindlasti ei kao arsti roll kuhugi. Üheski eelnevast andmestikust ja tehnoloogiast üksi ei ole kasu. Tehnoloogia ja arsti koostöös on võimalik vähendada ravieksimusi ja parandada teisi olulisi ravikvaliteedi näitajadi.
AI kindlasti muudab arstitööd. Võttes arstilt vajaduse sisestada andmeid digilukku, vabastades sellega peamist ressurssi, mida napib ja mille tõttu ~50% USA arstidest on läbipõlemise sümptomeid. See ressurss on aeg. Tänu sellele saavad arstid teha seda, milles me oleme head - olla inimesed. Olla empaatiline. Olla hetkes. Tunda kaasa. Kuulata. Anda patsiendile mõista, et mis ka ei juhtuks, on arst tema jaoks olemas ja annab endast kõik, et mitte ainult inimest terveks ravida, vaid vabastada teda ka kannatusest, mis haigusega kaasneb.
Ühiskonnana oleme leiutanud väga võimsa tööriista, mis on võimeline leidma tohututest andmemahtudest seaduspärasusi, milleks ükski inimene võimeline pole. Alustan lihtsamatest näidesest ja hakkan suurendama vajaminevaid andmemahtusid.
Aktiivsusmonitor detekteerib südame löögisageduse tõusu ilma, et inimese füüsiline aktiivsus oleks tõusnud. Palub koheselt teha inimesel EKG (elektrokardiogramm), detekteerib südame rütmihäire. EKG läheb arstile, patsient saab varakult arsti juurde. Arst määrab vajadusel ravimid.
Toidame algoritmi kümneid tuhandeid EKG-d, AI leiab seaduspärasusi, mille alusel on võimalus mitte-invasiivselt määrata inimese kaaliumi taset. Eluliselt oluline näitaja neerukahjustusega patsientideke.
Toidame närvivõrgustikku inimese DNA ning leiame seoseid geenide ja haiguste (nt autismi, diabeedi, Alzheimeri jt haiguste vahel) vahel.
Suurendame andmemahtusid veelgi. Toidame algoritmi andmeid, mis puudutavad: inimese sotsiaalseid ja käitumuslike mustreid, genoomikat jt 'oomikaid' (transkriptoomika, metaboloomika, proteoomika jt), immuunsüsteemi, mikrobioomi, anatoomi, keskkonna info, terviselugu, inimese kõne, biosensorite jt andmed, mis võimaldab meil saada väga hea ülevaate inimese tervise seisundist. Arvestades seda ravida haigust personaalsel tasemel. Kuid peamine, mis me selle andmestikuga teha saaksime, on muuta inimese käitumist, et vältida haiguste teket.
Kindlasti ei kao arsti roll kuhugi. Üheski eelnevast andmestikust ja tehnoloogiast üksi ei ole kasu. Tehnoloogia ja arsti koostöös on võimalik vähendada ravieksimusi ja parandada teisi olulisi ravikvaliteedi näitajadi.
AI kindlasti muudab arstitööd. Võttes arstilt vajaduse sisestada andmeid digilukku, vabastades sellega peamist ressurssi, mida napib ja mille tõttu ~50% USA arstidest on läbipõlemise sümptomeid. See ressurss on aeg. Tänu sellele saavad arstid teha seda, milles me oleme head - olla inimesed. Olla empaatiline. Olla hetkes. Tunda kaasa. Kuulata. Anda patsiendile mõista, et mis ka ei juhtuks, on arst tema jaoks olemas ja annab endast kõik, et mitte ainult inimest terveks ravida, vaid vabastada teda ka kannatusest, mis haigusega kaasneb.
May 23, 2019
A very up to date and well researched book on the world of AI and its application in medicine.
For someone with little knowledge about AI like myself, it gives a quick overview of AI, in what sectors AI tools are being used, and then proceeds to talk about its application within different branches of medicine, but also within other related areas such as drug discovery, omics and mental health.
At the same time he also touches on issues of empathy in medicine and the patient-doctor relationship.
"...we can choose a technological solution to the profound human disconnection that exists today in healthcare..."
For someone with little knowledge about AI like myself, it gives a quick overview of AI, in what sectors AI tools are being used, and then proceeds to talk about its application within different branches of medicine, but also within other related areas such as drug discovery, omics and mental health.
At the same time he also touches on issues of empathy in medicine and the patient-doctor relationship.
"...we can choose a technological solution to the profound human disconnection that exists today in healthcare..."
June 14, 2020
With vast mountains of data generated every day, the medical profession is ripe for the application of deep learning to transform its very landscape.
From bringing precision medicine to life to improving the utility of electronic health records systems, the potential of AI to improve medicine for doctors and patients alike is extraordinary.
The potential disruptive changes this might bring are also considered in the book, particularly with regards to the cries that we should stop training all radiologists now - Topol suggests that new generations of radiologists will work more as "scan specialists," who work in tandem with AI to read pathological slides as well as MRIs, X-rays and CT scans.
Unfortunately, the book feels somewhat choppy at times, feeling more like a collection of essays than a cohesive book. However, I don't think that this is a significant problem; the separation allows Topol to compartmentalise his thesis into its various subsections quite effectively.
He reminds us that AI may not just be able to build upon what is good and can be made excellent; it can also help to fix the "problem areas" in medicine that have crept in over the years, in particular, time spent with patients. Topol suggests that AI could help to maximise doctor patient interactions, by streamlining a lot of the various administrative processes that a doctor might have to attend to during a consultation.
It's a pertinent reminder that medicine is inherently a human enterprise, and that should not be forgotten in the pursuit of optimisation with deep learning.
"To cure sometimes, to relieve often, to comfort always"
From bringing precision medicine to life to improving the utility of electronic health records systems, the potential of AI to improve medicine for doctors and patients alike is extraordinary.
The potential disruptive changes this might bring are also considered in the book, particularly with regards to the cries that we should stop training all radiologists now - Topol suggests that new generations of radiologists will work more as "scan specialists," who work in tandem with AI to read pathological slides as well as MRIs, X-rays and CT scans.
Unfortunately, the book feels somewhat choppy at times, feeling more like a collection of essays than a cohesive book. However, I don't think that this is a significant problem; the separation allows Topol to compartmentalise his thesis into its various subsections quite effectively.
He reminds us that AI may not just be able to build upon what is good and can be made excellent; it can also help to fix the "problem areas" in medicine that have crept in over the years, in particular, time spent with patients. Topol suggests that AI could help to maximise doctor patient interactions, by streamlining a lot of the various administrative processes that a doctor might have to attend to during a consultation.
It's a pertinent reminder that medicine is inherently a human enterprise, and that should not be forgotten in the pursuit of optimisation with deep learning.
"To cure sometimes, to relieve often, to comfort always"
April 12, 2020
Eric Topol does a great job giving a broad overview of AI and the implications for healthcare. I appreciated that he was able to talk broadly whilst also giving specific examples. I also liked that I did not need to know much about artificial intelligence.
He starts by explaining the issues with the healthcare systems of today, which he sums up with the saying ‘shallow medicine’ - insufficient time, connection, data, and context in our consultations. He then describes artificial intelligence and gives examples of the progress AI is making is specific areas of healthcare (which is too broad to start listing).
He finishes with the idea that AI will undoubtedly change healthcare in the coming future, but will not replace the practitioner. AI in healthcare can be used to either increase the productivity and efficiency of healthcare practitioners, or it can be used to give us more time with our patients. He emphasises the importance of actively using this as an opportunity to make healthcare human again - spending more time with our patients, touching them, being empathetic, and connecting with them.
Key things that I took away from this book are:
- We have to actively fight to ensure AI helps us connect more with our patients (i.e. have more time to be truly present in consultations), rather than just making us more efficient and productive.
- Completely autonomous AI doctors will likely never occur, but it will improve healthcare in incredible ways.
- It is going to become increasingly important for healthcare professionals to have an understanding of AI.
He starts by explaining the issues with the healthcare systems of today, which he sums up with the saying ‘shallow medicine’ - insufficient time, connection, data, and context in our consultations. He then describes artificial intelligence and gives examples of the progress AI is making is specific areas of healthcare (which is too broad to start listing).
He finishes with the idea that AI will undoubtedly change healthcare in the coming future, but will not replace the practitioner. AI in healthcare can be used to either increase the productivity and efficiency of healthcare practitioners, or it can be used to give us more time with our patients. He emphasises the importance of actively using this as an opportunity to make healthcare human again - spending more time with our patients, touching them, being empathetic, and connecting with them.
Key things that I took away from this book are:
- We have to actively fight to ensure AI helps us connect more with our patients (i.e. have more time to be truly present in consultations), rather than just making us more efficient and productive.
- Completely autonomous AI doctors will likely never occur, but it will improve healthcare in incredible ways.
- It is going to become increasingly important for healthcare professionals to have an understanding of AI.
April 29, 2019
Well researched, eye-opening and fascinating. AI is the future of medicine and I think we need to embrace it and make sure that patients get the most out of it.
January 11, 2023
I really enjoyed this book. Five months ago I started working in healthcare AI, with zero background in AI/ML, (but fortunately I don't need a background in AI for my role, as I'm not on the tech side). I'm learning a lot about it naturally through working around AI engineers, physicians, researchers, IT folks, data scientists, etc., but I'm trying to accelerate my learning as much as possible. I picked up this book after someone in a meeting mentioned it. It's wonderful. Topol does an amazing job at breaking down really complex concepts and explaining it in very easy-to-understand ways so it becomes extremely accessible to the reader, while also showing why robots won't take over all our jobs. I also geeked out a little bit when he mentioned some medical education work (like the role of humanities and art appreciation in diagnosis), because that has been a significant area of focus for me in my PhD work. So, I appreciated the diversity of experiences within medicine that he managed to work into his narrative.
It is interesting to see even in the last few years how much progress has been made, too, as some of the things he mentions in the book have already evolved rapidly. Super fascinating. This is the book I'll give someone if they ask me what I do for a living.
It is interesting to see even in the last few years how much progress has been made, too, as some of the things he mentions in the book have already evolved rapidly. Super fascinating. This is the book I'll give someone if they ask me what I do for a living.
June 29, 2022
This might have been good as a guardian long read. The general theme is interesting, but the book is far too full of case studies, and disappointingly light on analysis of the actual implications of greater use of AI in healthcare, particularly regarding the ethical dimensions of the issue. Ultimately, unlike Topol, I don't believe we can rely on unproven technology to save us. We have to put in the work ourselves.
May 30, 2019
One way to describe modern medicine is "shallow". This book explores how AI would actually deepen in and how we should not be afraid of what it can bring to the table.
October 11, 2022
با تشکر از آرش عزیز که کتاب صوتی این فایل رو برام گیر آورد و بهم اجازه داد صرفا با پلی کردن و گوش ندادن بهش، عذاب وجدان نخوندن این کتاب رو از گردنم بردارم.3>
It’s no wonder that we have an opioid epidemic when it’s a lot quicker and easier for doctors to prescribe narcotics than to listen to and understand patients.
اولین بار و بدون هیچ مطالعهای با شنیدن کلمهی AI یاد فیلم Elysium افتادم. ماشینی که هربیماریای رو درمان میکنه و براش فرقی نداره که داستان چیه. کنسر؟ مشکلی نداره. پسوریازیس؟ درمان میکنیم. آب حوض؟ میکشیم.
هرچیزی که مرتبط با صنعت روباتیک و هوش مصنوعی باشه این روزا این جمله رو توی ذهن مردم زنده میکنه که «به زودی روباتهای جای پزشکا رو میگیرن.» اما این فعلا در حد یه شوخیه. پروتکلها و سیستمهایی که برای هوش مصنوعی طراحی شدن خیلی محدودتر از این حرفها هستن و برای طراحی هرکدوم مدت زمان بسیار زیادی لازمه تا نتیجهای بهتر از نمونهی انسانی به دست بیاد. اینطوری نیست که یه نفر بشینه، دیتاهای مربوط به دیابت رو وارد یه سیستم کنه، به اون سیستم آموزش بده و تهش بگه خب، دیگه کسی لازم نیست در زمینهی دیابت به پزشک مراجعه کنه و فقط با این سیستم همهی کاراش راه میفته. همینکه یه نفر بتونه فقط و فقط برای تشخیص دیابت و اونم تو یه جمعیت با ویژگی خاص یه شبکه طراحی کنه خیلی هنر کرده. چیزی که امروزه داریم ترکیبی از همکاری هوش مصنوعی و پزشکان انسانیه که Deep medicine نامیده میشه. پزشکی عمیق قادره که هر انسان رو به صورت منحصربفرد و جداگانه، براساس تمام اطلاعات ژنومی مربوط به اون بررسی کنه. در مرحلهی دوم با شناسایی الگوها و یادگیری عمیق این اطلاعات شخصی رو در یک مسیر صحیح برای شناختن درمان یا تشخیص مناسب قرار میده و در بخش سوم با ایجاد یک امپاتی صحیح، رابطهی مناسبی رو بین پزشک و بیمار برقرار میکنه. پس همهچیز اونقدرها هم قرار نیست روباتیک باشه=))).
patients exist in a world of insufficient data, insufficient time, insufficient context, and insufficient presence. Or, as I say, a world of shallow medicine.
درمقابل پزشکی عمیق، معضلی به اسم پزشکی سطحی حتی مدعیان پیشرفت در خدمات درمان و سلامت رو درگیر خودش کرده. اینکه به صورت سطحی و زودگذر علائم بیمار رو مرتبط با یک بیماری بدونیم باعث میشه سالانه میلیونها دلار از بودجهی بیماران و سیستم بهداشتی صرف اقدامات تشخیصی و درمانی نامناسبی بشه که هیچ نیازی بهشون نیست و استهلاک ماشینهای تشخیصی و کمبود برخی کیتها توی همین مملکت خودمون این معضل رو چندبرابر بزرگتر میکنه. یک یافتهی تشخیصی غیرمرتبط میتونه بار روانی و مالی بزرگی رو به بیمار تحمیل کنه:
Isaac Kohane nicknamed them, “incidentalomas.”
To be a good diagnostician, a physician needs to acquire a large set of labels for diseases, each of which binds an idea of the illness and its symptoms, possible antecedents and causes, possible developments and consequences, and possible interventions to cure or mitigate the illness.
—DANIEL KAHNEMAN
طبق مدل کانمن ما دو مدل فکری داریم، یکی سریع و بدون فکر و دیگری آهسته و آگاهانه. کانمن پیشنهاد میکنه هروقت خواستیم یه اندیشهی درست داشته باشیم باید بعد از یه تفکر بدون درنگ، یکمی صبر کنیم، برگردیم و با خودمون فکر کنیم، غیر از این چی میتونه باشه.
بخشی از مشکل سیستم درمانی نگرفتن فیدبک مناسبه، بیماری که درمان میگیره، معمولا بدون بازخورد مرخص میشه و معلوم نمیشه که آیا درمان مناسب بوده یا نه. همین نبودن بازخورد مناسب در گذر زمان باعث افزایش اعتمادبهنفس میشه. از طرفی دیگه ما با خطایی به اسم «در دسترس بودن» سر و کار داریم. اگر در طول روز به بیماری با تشخیص نادر بربخوریم، احتمال اینکه در طول روز تشخیصهای مبتنی بر اون بیماری بذاریم افزایش پیدا میکنه. پلتفرمهای مختلفی برای تشخیص و درمان طراحی شدن. مشکلی که در رابطه با اینها وجود داره اینه که علائم بیمار همیشه صفر و یکی نیستن. ممکنه حس درد رو تحت عنوان فشار بیان کنه یا از زبان بدنش برای بیان مشکلش استفاده کنه.
A computer reading a scientific paper requires human oversight to pick out key words and findings.
بخشی از پزشکان، یعنی کسانی که رادیولوژیست، پاتولوژیست و درماتولوژیست نیستن رو میشه تحت عنوان پزشکان بدون Pattern نامگذاری کرد. کار اینا کمی متفاوته و با الگوریتمهای متعددی سروکار داره که مشارکت هوش مصنوعی میتونه درکنار پزشک، محیط مناسبی رو برای بیکار فراهم کنه.
We have trapped ourselves in a binary world of data interpretation—normal or abnormal—and are ignoring rich, granular, and continuous data that we could be taking advantage of.
باید توجه داشت که امروزه خیلی چیزها به صورت صفر و یکی بیان میشن. خیلی از نرمالهایی که نوشته میشن چندان به اختصاصیت راجع بهشون فکر نشده. مثلا نژاد و شرایط زندگی فرد درشون دخیل دونسته نشده. وجود سیستمی که بتونه تمام اطلاعات فرد رو جمعآوری و متمرکز کنه میتونه به پزشک اجازه بده که زمان بیشتری رو صرف شنیدن حرفای مراجعانش بکنه.
از یه جایی به بعد دیگه نوت برداشتن از کتاب رو متوقف کردم چون پروسهی خوندنش داشت خیلی فرسایشی میشد و زمان زیادی ازم گرفت. کتاب پره از مثالهای متفاوتی از مشارکت هوش مصنوعی توی حیطههای مختلفی از زندگی ولی نوشتار و متن چندان باب میلم نبود و فقط خوشحالم که تمومش کردم ولی احساس میکنم از همین اولش، این حیطه چیزی نیست که دلم بخواد توش نقشی داشته باشم. فعلا دلم میخواد چندتا کتاب بوجی موجی و بدون اساس علمی بخونم.
It’s no wonder that we have an opioid epidemic when it’s a lot quicker and easier for doctors to prescribe narcotics than to listen to and understand patients.
اولین بار و بدون هیچ مطالعهای با شنیدن کلمهی AI یاد فیلم Elysium افتادم. ماشینی که هربیماریای رو درمان میکنه و براش فرقی نداره که داستان چیه. کنسر؟ مشکلی نداره. پسوریازیس؟ درمان میکنیم. آب حوض؟ میکشیم.
هرچیزی که مرتبط با صنعت روباتیک و هوش مصنوعی باشه این روزا این جمله رو توی ذهن مردم زنده میکنه که «به زودی روباتهای جای پزشکا رو میگیرن.» اما این فعلا در حد یه شوخیه. پروتکلها و سیستمهایی که برای هوش مصنوعی طراحی شدن خیلی محدودتر از این حرفها هستن و برای طراحی هرکدوم مدت زمان بسیار زیادی لازمه تا نتیجهای بهتر از نمونهی انسانی به دست بیاد. اینطوری نیست که یه نفر بشینه، دیتاهای مربوط به دیابت رو وارد یه سیستم کنه، به اون سیستم آموزش بده و تهش بگه خب، دیگه کسی لازم نیست در زمینهی دیابت به پزشک مراجعه کنه و فقط با این سیستم همهی کاراش راه میفته. همینکه یه نفر بتونه فقط و فقط برای تشخیص دیابت و اونم تو یه جمعیت با ویژگی خاص یه شبکه طراحی کنه خیلی هنر کرده. چیزی که امروزه داریم ترکیبی از همکاری هوش مصنوعی و پزشکان انسانیه که Deep medicine نامیده میشه. پزشکی عمیق قادره که هر انسان رو به صورت منحصربفرد و جداگانه، براساس تمام اطلاعات ژنومی مربوط به اون بررسی کنه. در مرحلهی دوم با شناسایی الگوها و یادگیری عمیق این اطلاعات شخصی رو در یک مسیر صحیح برای شناختن درمان یا تشخیص مناسب قرار میده و در بخش سوم با ایجاد یک امپاتی صحیح، رابطهی مناسبی رو بین پزشک و بیمار برقرار میکنه. پس همهچیز اونقدرها هم قرار نیست روباتیک باشه=))).
patients exist in a world of insufficient data, insufficient time, insufficient context, and insufficient presence. Or, as I say, a world of shallow medicine.
درمقابل پزشکی عمیق، معضلی به اسم پزشکی سطحی حتی مدعیان پیشرفت در خدمات درمان و سلامت رو درگیر خودش کرده. اینکه به صورت سطحی و زودگذر علائم بیمار رو مرتبط با یک بیماری بدونیم باعث میشه سالانه میلیونها دلار از بودجهی بیماران و سیستم بهداشتی صرف اقدامات تشخیصی و درمانی نامناسبی بشه که هیچ نیازی بهشون نیست و استهلاک ماشینهای تشخیصی و کمبود برخی کیتها توی همین مملکت خودمون این معضل رو چندبرابر بزرگتر میکنه. یک یافتهی تشخیصی غیرمرتبط میتونه بار روانی و مالی بزرگی رو به بیمار تحمیل کنه:
Isaac Kohane nicknamed them, “incidentalomas.”
To be a good diagnostician, a physician needs to acquire a large set of labels for diseases, each of which binds an idea of the illness and its symptoms, possible antecedents and causes, possible developments and consequences, and possible interventions to cure or mitigate the illness.
—DANIEL KAHNEMAN
طبق مدل کانمن ما دو مدل فکری داریم، یکی سریع و بدون فکر و دیگری آهسته و آگاهانه. کانمن پیشنهاد میکنه هروقت خواستیم یه اندیشهی درست داشته باشیم باید بعد از یه تفکر بدون درنگ، یکمی صبر کنیم، برگردیم و با خودمون فکر کنیم، غیر از این چی میتونه باشه.
بخشی از مشکل سیستم درمانی نگرفتن فیدبک مناسبه، بیماری که درمان میگیره، معمولا بدون بازخورد مرخص میشه و معلوم نمیشه که آیا درمان مناسب بوده یا نه. همین نبودن بازخورد مناسب در گذر زمان باعث افزایش اعتمادبهنفس میشه. از طرفی دیگه ما با خطایی به اسم «در دسترس بودن» سر و کار داریم. اگر در طول روز به بیماری با تشخیص نادر بربخوریم، احتمال اینکه در طول روز تشخیصهای مبتنی بر اون بیماری بذاریم افزایش پیدا میکنه. پلتفرمهای مختلفی برای تشخیص و درمان طراحی شدن. مشکلی که در رابطه با اینها وجود داره اینه که علائم بیمار همیشه صفر و یکی نیستن. ممکنه حس درد رو تحت عنوان فشار بیان کنه یا از زبان بدنش برای بیان مشکلش استفاده کنه.
A computer reading a scientific paper requires human oversight to pick out key words and findings.
بخشی از پزشکان، یعنی کسانی که رادیولوژیست، پاتولوژیست و درماتولوژیست نیستن رو میشه تحت عنوان پزشکان بدون Pattern نامگذاری کرد. کار اینا کمی متفاوته و با الگوریتمهای متعددی سروکار داره که مشارکت هوش مصنوعی میتونه درکنار پزشک، محیط مناسبی رو برای بیکار فراهم کنه.
We have trapped ourselves in a binary world of data interpretation—normal or abnormal—and are ignoring rich, granular, and continuous data that we could be taking advantage of.
باید توجه داشت که امروزه خیلی چیزها به صورت صفر و یکی بیان میشن. خیلی از نرمالهایی که نوشته میشن چندان به اختصاصیت راجع بهشون فکر نشده. مثلا نژاد و شرایط زندگی فرد درشون دخیل دونسته نشده. وجود سیستمی که بتونه تمام اطلاعات فرد رو جمعآوری و متمرکز کنه میتونه به پزشک اجازه بده که زمان بیشتری رو صرف شنیدن حرفای مراجعانش بکنه.
از یه جایی به بعد دیگه نوت برداشتن از کتاب رو متوقف کردم چون پروسهی خوندنش داشت خیلی فرسایشی میشد و زمان زیادی ازم گرفت. کتاب پره از مثالهای متفاوتی از مشارکت هوش مصنوعی توی حیطههای مختلفی از زندگی ولی نوشتار و متن چندان باب میلم نبود و فقط خوشحالم که تمومش کردم ولی احساس میکنم از همین اولش، این حیطه چیزی نیست که دلم بخواد توش نقشی داشته باشم. فعلا دلم میخواد چندتا کتاب بوجی موجی و بدون اساس علمی بخونم.
January 25, 2022
Erneut eine beeindruckend erarbeitete Vision zur Zukunft des Gesundheitswesens. Dieses Mal darauf bezogen wie künstliche Intelligenz, Big Data, und Deep Learning für bessere Outcomes, Prävention und vor allem Doktor-Patient-Relationships führen kann.
Dieses Buch ist mit weniger Anekdoten und Fallbeispielen ausgestattet, diese wurden dafür durch vergleichende Studien (viele Zahlen - langweilig zu lesen) und einer schlüssigeren Narrative ersetzt. Hier wird sich auch kritischer mit dem Sachverhalt auseinandergesetzt als es bei seinem vorigen Buch der Fall war, wodurch man ein ganzheitlicheres Bild erhält und letztendlich seiner Schlussfolgerung zustimmen muss.
Auch diese Vision ist geprägt von einem fast träumerischen, realitätsfernen (weil das Gesundheitssystem so gefickt ist) Optimismus. Hier wird ein klarer Weg vorgezeigt, wie man von kalten, gefühllosen Doktor_innen, zu empathischen, gefühlvollen Heiler_innen kommen kann. Künstliche Intelligenz kann es schaffen wieder mehr Wert auf das zwischenmenschliche beim Arztbesuch zu legen.
Wieder einmal Pflichtlektüre für alle Mediziner_innen und Arbeiter_innen des Gesundheitswesens!
Danke Eric
8/10 📖
Dieses Buch ist mit weniger Anekdoten und Fallbeispielen ausgestattet, diese wurden dafür durch vergleichende Studien (viele Zahlen - langweilig zu lesen) und einer schlüssigeren Narrative ersetzt. Hier wird sich auch kritischer mit dem Sachverhalt auseinandergesetzt als es bei seinem vorigen Buch der Fall war, wodurch man ein ganzheitlicheres Bild erhält und letztendlich seiner Schlussfolgerung zustimmen muss.
Auch diese Vision ist geprägt von einem fast träumerischen, realitätsfernen (weil das Gesundheitssystem so gefickt ist) Optimismus. Hier wird ein klarer Weg vorgezeigt, wie man von kalten, gefühllosen Doktor_innen, zu empathischen, gefühlvollen Heiler_innen kommen kann. Künstliche Intelligenz kann es schaffen wieder mehr Wert auf das zwischenmenschliche beim Arztbesuch zu legen.
Wieder einmal Pflichtlektüre für alle Mediziner_innen und Arbeiter_innen des Gesundheitswesens!
Danke Eric
8/10 📖
December 15, 2020
Initially picked this up just to get a better context on where medicine is going, but ended up learning a little more, about the progression of AI and even Topol’s reflections on being a personal doctor.
Read
June 5, 2025I didn't finish. I was up against time constraints and had six audio books borrowed. It was easier to return this book and catch a breath then to power through. I wasn't convinced of Dr. Topol's arguments.
But I am eager to read his book Super Agers.
But I am eager to read his book Super Agers.
January 23, 2022
Really interesting, full of deep thought about AI in medicine.
April 1, 2020
Ultimately, this is an optimistic book. "The greatest opportunity offered by AI is not reducing errors or workloads, or even curing cancer: it is the opportunity to restore the precious and time-honored connection and trust—the human touch—between patients and doctors."
There's detail on what machine learning can do right now. That includes diagnosis from radiography, virutal medicine (telemedicine and chatbots), personalisation of healthcare, diet, mental health, and AI as a tool for use by the clinician.
I found it astonishing how much progress has been made. Yet: "We're still in the earliest days of AI in medicine. The field is long on computer algorithmic validation and promises but very short on real-world, clinical proof of effectiveness." Nevertheless, "it is inevitable that narrow AI [specific, targetting algorithms] will take hold".
My takeaway from the book was the hope for "deep empathy": knowing about the patient, their history, and having the time to use that knowledge. "It's our chance, perhaps the ultimate one, to bring back real medicine: Presence. Empathy. Trust. Caring. Being Human."
There's detail on what machine learning can do right now. That includes diagnosis from radiography, virutal medicine (telemedicine and chatbots), personalisation of healthcare, diet, mental health, and AI as a tool for use by the clinician.
I found it astonishing how much progress has been made. Yet: "We're still in the earliest days of AI in medicine. The field is long on computer algorithmic validation and promises but very short on real-world, clinical proof of effectiveness." Nevertheless, "it is inevitable that narrow AI [specific, targetting algorithms] will take hold".
My takeaway from the book was the hope for "deep empathy": knowing about the patient, their history, and having the time to use that knowledge. "It's our chance, perhaps the ultimate one, to bring back real medicine: Presence. Empathy. Trust. Caring. Being Human."
December 30, 2020
Topol has written a sweeping and well-rounded book about AI in medicine today. This book is for everyone currently in the healthcare system or working in the field of AI. It is also a great book for someone that generally wants to learn more about the intersection of the two.
I found Topols writing and argumentation easy to follow and the book balances the intriguing possibilities of new AI methods in medicine with the current (lack of) clinical trials and testing.
The breadth of the book, touching upon everything from care, drug discovery, doctor productivity and insurance - show just how broadly AI-impact will be in the area of healthcare.
I liked to book and would recommend to anyone interested in the general topic. If you have read this far in the review, you probably should read the book also :)
I found Topols writing and argumentation easy to follow and the book balances the intriguing possibilities of new AI methods in medicine with the current (lack of) clinical trials and testing.
The breadth of the book, touching upon everything from care, drug discovery, doctor productivity and insurance - show just how broadly AI-impact will be in the area of healthcare.
I liked to book and would recommend to anyone interested in the general topic. If you have read this far in the review, you probably should read the book also :)
November 12, 2021
Embarradita del actual uso de la inteligencia artificial en la medicina, en especial la muy breve mención de su uso en la Patología.
Igual y el libro pudo resumirse en un bueno y largo review, o en un tomo más corto y conciso.
Igual y el libro pudo resumirse en un bueno y largo review, o en un tomo más corto y conciso.
March 28, 2021
Two stars for me -- this was okay, but that may simply reflect that this seems to be intended as a high-level overview of AI and similar technologies and their current and future effects on the practice of medicine and healthcare. As someone who works in health IT and follows issues related to AI, machine learning, the economics of heath care, and so on, there was very little in this book that was new to me. If you're looking for a high-level overview of AI and healthcare, this is a good place to start.
I do like Topol's emphasis on having AI, machine learning, and so on, be used to allow doctors to do a better job connecting with patients. It's become obvious in recent years that so-called "social determinants of health" are in many ways much better predictors of health and illness -- in other words, knowing where someone lives and what kind of family relationships they have is, in many cases, more useful than a large battery of fancy lab tests. Perhaps in a similar way, AI can allow us to reconceptualize the doctor-patient relationship as more of a trusted, close advisor/friend.
I do wonder about how much AI and other technology can really enable Topol's "deep empathy" . I see two problems: Baumol's cost disease and regular capitalistic incentives.
First, there's Baumol's cost disease. It seems that doctors and patients would prefer, say, 30 minute visits. But the cost disease argument is that if you have something defined by a fixed amount of time, its cost must rise. Yes, algorithms and machine learning diagnostic tools can improve the care, but if providers and patients want that 30 minutes, that cost must necessarily rise, and you can't get the sort of order-of-magnitude productivity gains that you need to really change costs.
The second thing relates to the larger economic system driven by profit or revenue. Say you've got your 30-minute visits. Doctors and patients like it. Say doctors have 8 hours, or 16 patients, of time a day. But the healthcare system leadership thinks "if we make those visits 25 minutes, the doctors can see 19 patients a day! That's over 18% more patients every day!" But the same logic keeps applying, and before long we're back with the ultra-short rushed visits we have now.
This isn't necessarily about moustache-twirling predatory capitalists; even government-driven programs like Medicare or the NHS can apply this kind of pressure. In addition to longer visits, though, patients also want better access, both for scheduling convenience and medical necessity -- no one wants to wait to see the doctor if you are sick and want advice, but not so sick that emergenccy or urgent care is right.
b
I do like Topol's emphasis on having AI, machine learning, and so on, be used to allow doctors to do a better job connecting with patients. It's become obvious in recent years that so-called "social determinants of health" are in many ways much better predictors of health and illness -- in other words, knowing where someone lives and what kind of family relationships they have is, in many cases, more useful than a large battery of fancy lab tests. Perhaps in a similar way, AI can allow us to reconceptualize the doctor-patient relationship as more of a trusted, close advisor/friend.
I do wonder about how much AI and other technology can really enable Topol's "deep empathy" . I see two problems: Baumol's cost disease and regular capitalistic incentives.
First, there's Baumol's cost disease. It seems that doctors and patients would prefer, say, 30 minute visits. But the cost disease argument is that if you have something defined by a fixed amount of time, its cost must rise. Yes, algorithms and machine learning diagnostic tools can improve the care, but if providers and patients want that 30 minutes, that cost must necessarily rise, and you can't get the sort of order-of-magnitude productivity gains that you need to really change costs.
The second thing relates to the larger economic system driven by profit or revenue. Say you've got your 30-minute visits. Doctors and patients like it. Say doctors have 8 hours, or 16 patients, of time a day. But the healthcare system leadership thinks "if we make those visits 25 minutes, the doctors can see 19 patients a day! That's over 18% more patients every day!" But the same logic keeps applying, and before long we're back with the ultra-short rushed visits we have now.
This isn't necessarily about moustache-twirling predatory capitalists; even government-driven programs like Medicare or the NHS can apply this kind of pressure. In addition to longer visits, though, patients also want better access, both for scheduling convenience and medical necessity -- no one wants to wait to see the doctor if you are sick and want advice, but not so sick that emergenccy or urgent care is right.
b
March 27, 2019
Eric Topol is always interesting be it a book, an interview, a podcast or on Twitter. On Twitter to me he wins the MVT award for most valuable tweeter.
This is I believe the third book of his that I’ve read. I am not a novice to AI/ML and I lecture on these topics.
This is a very thorough book. Almost every page is filled with information worth checking out even further. It covers the current state of AI in medicine thoroughly. From journal articles to companies carrying out the efforts I can’t see that there are omissions.
Visions of the future are usually speculative. The vision of a speech recognition system assisting in writing a chart to help ease the burden of EHR is something I’ve lectured to biomedical engineering students for several years. Some day will likely happen.
He speaks of AI/ML as being an assist to the medical field rather than a replacement. This is starting to happen and should be in practice more and more before long. The systems as assistant and not as replacement certainly makes a lot of sense.
He is more optimistic than I that corporate medicine will allow doctors to spend their time demonstrating empathy and perhaps longer than 12 minutes? Dr. Topol also expresses the need to select medical students more on empathy and humanistic characteristics. But several paragraphs later he said that we need more physicians to understand algorithm construction and be technical AI/ML experts. We need both but it is hard to be both. And while you can look things up - you need your medical database and experience in your head - not on google.
But overall - I do not hesitate to give this book a five star rating. It gives a view of the power of AI/ML, where we are now and provides a vision of an idealistic possible future. You will not learn how to do machine learning or have an outline of where to begin. But that is not the task of
this text. There are many courses available online for that: start with University of Washington and consider Andrew Ng at Stanford or several other institutions through MOOCs or degree programs.
This is I believe the third book of his that I’ve read. I am not a novice to AI/ML and I lecture on these topics.
This is a very thorough book. Almost every page is filled with information worth checking out even further. It covers the current state of AI in medicine thoroughly. From journal articles to companies carrying out the efforts I can’t see that there are omissions.
Visions of the future are usually speculative. The vision of a speech recognition system assisting in writing a chart to help ease the burden of EHR is something I’ve lectured to biomedical engineering students for several years. Some day will likely happen.
He speaks of AI/ML as being an assist to the medical field rather than a replacement. This is starting to happen and should be in practice more and more before long. The systems as assistant and not as replacement certainly makes a lot of sense.
He is more optimistic than I that corporate medicine will allow doctors to spend their time demonstrating empathy and perhaps longer than 12 minutes? Dr. Topol also expresses the need to select medical students more on empathy and humanistic characteristics. But several paragraphs later he said that we need more physicians to understand algorithm construction and be technical AI/ML experts. We need both but it is hard to be both. And while you can look things up - you need your medical database and experience in your head - not on google.
But overall - I do not hesitate to give this book a five star rating. It gives a view of the power of AI/ML, where we are now and provides a vision of an idealistic possible future. You will not learn how to do machine learning or have an outline of where to begin. But that is not the task of
this text. There are many courses available online for that: start with University of Washington and consider Andrew Ng at Stanford or several other institutions through MOOCs or degree programs.
Want to read
March 12, 2020Nice review at Science magazine: https://blogs.sciencemag.org/books/20...
Excerpt:
"Eric Topol ... is optimistic about the future of health care. In Deep Medicine, he anticipates that new machine learning technologies will improve the precision and accuracy of disease diagnosis, thus providing a better way to identify the best therapies. ... He hopes that the time freed up by these approaches will be devoted to reviving humane medical practices. ...
The great contribution of this book is that Topol synthesizes the fragmentary views that we who work in this field gain from day-to-day reading into a cohesive vision of a future in which medical care is about human care. Alas, achieving that depends on much more than improved technological support for clinical medicine. Hopefully, the economic and administrative forces that have done much to frustrate other recent visionaries will not derail this new plan."
-- The reviewer is head of the Clinical Decision-Making Group, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL)
Excerpt:
"Eric Topol ... is optimistic about the future of health care. In Deep Medicine, he anticipates that new machine learning technologies will improve the precision and accuracy of disease diagnosis, thus providing a better way to identify the best therapies. ... He hopes that the time freed up by these approaches will be devoted to reviving humane medical practices. ...
The great contribution of this book is that Topol synthesizes the fragmentary views that we who work in this field gain from day-to-day reading into a cohesive vision of a future in which medical care is about human care. Alas, achieving that depends on much more than improved technological support for clinical medicine. Hopefully, the economic and administrative forces that have done much to frustrate other recent visionaries will not derail this new plan."
-- The reviewer is head of the Clinical Decision-Making Group, MIT Computer Science and Artificial Intelligence Laboratory (CSAIL)
January 25, 2022
Topol probably doesn't get much of what Medicine is. He does have a diploma that certifies he has spent the tuition money for the required number of Med School years. And he has a license that makes him a member of a very nasty guild. Not much else.
Topol gets less about healthcare. Of course, like any entitled white male with a higher than average wage, he is convinced he is the angel sent from above to fix the system.
Now, AI, that is something Topol gets less than the other two. Never mind. For him AI is yet another god that will help the enlightened white males like himself bring Socialist utopia to the imbecile masses.
As for Human, that is simply an emotional term, devoid of reason, like most of Topol's collection of easy fallacies. But yes, the poor guy lives in a world resembling much the world of the cave man: Medicine, AI, Healthcare, all gods walking the Earth, and only demi-gods from his guild might interact with the said gods.
Topol gets less about healthcare. Of course, like any entitled white male with a higher than average wage, he is convinced he is the angel sent from above to fix the system.
Now, AI, that is something Topol gets less than the other two. Never mind. For him AI is yet another god that will help the enlightened white males like himself bring Socialist utopia to the imbecile masses.
As for Human, that is simply an emotional term, devoid of reason, like most of Topol's collection of easy fallacies. But yes, the poor guy lives in a world resembling much the world of the cave man: Medicine, AI, Healthcare, all gods walking the Earth, and only demi-gods from his guild might interact with the said gods.
July 16, 2020
In Deep Medicine Eric Topol gives an excellent snapshot of the current state of AI in healthcare. While this book gives a comprehensive current assessment of AI in medicine, Topol spends little time looking into the future or exploring the philosophical implications of this technology. Given the title of this book, I was hoping for more of this future oriented analysis. Still giving it 4 stars as it was well written, well researched, and at least gave cursory thought to the future of AI in healthcare.
May 13, 2022
Informative book, it gives a bird's eye overview of the current applications and potential of the use of AI in medicine and how it can be successfully employed to aid human experience and judgement. The book walks through use cases and real events while setting the ground for a future expansion of the field. It is a very useful read not only for those interested in AI in general but also for everyone who wants to learn more about where part of the progress in the medicine and life sciences fields is heading.
December 23, 2022
Disappointing. Lots of chatter without much meat
March 22, 2025
This book has been on my list to read for a while and I was worried that it even though it was published in 2019 that it might already feel dated. I did not find that to be the case. I finally read it after returning from an educational conference where incorporating AI into teaching about psychiatry was the main theme. Talk after talk about how we could use AI to teach somewhat worse than we are capable (shoddy simulations, clunky patient scenarios). I had also heard a lot about AI chat bots for therapy,or more about how they could replace humans.
But I enjoyed this book because it was my first deep look at how AI could be used not to replace humans in medicine, but to do things humans are not capable of doing. Sifting through huge amounts of genomic or personal data to identify patterns. Reorganizing our conception of mental illness (instead of a simplistic symptom based frame work we have now). My favorite fact; the total number of drug-like molecules (small organic molecules that have a low molecular weight, soluble in both water and fat) numbers 10^60, far more than the total number of atoms in the solar system.
Unfavorable outcomes are certainly possible with AI. Profit motives will likely force implementation of products that do better or only slightly worse than humans. But I appreciated the positive look on what medicine could look like with some of the benefits of AI.
But I enjoyed this book because it was my first deep look at how AI could be used not to replace humans in medicine, but to do things humans are not capable of doing. Sifting through huge amounts of genomic or personal data to identify patterns. Reorganizing our conception of mental illness (instead of a simplistic symptom based frame work we have now). My favorite fact; the total number of drug-like molecules (small organic molecules that have a low molecular weight, soluble in both water and fat) numbers 10^60, far more than the total number of atoms in the solar system.
Unfavorable outcomes are certainly possible with AI. Profit motives will likely force implementation of products that do better or only slightly worse than humans. But I appreciated the positive look on what medicine could look like with some of the benefits of AI.
December 28, 2024
Kuulasin audioraamatuna, ilmselt muidu poleks suutnud lõpuni jõuda. Samas annan 4 punkti, sest raamat oli iseenesest kaasahaarav ja huvitav. Ma ei teadnud, et nii palju erinevaid meditsiinilisi AI rakendusi ja lahendusi on juba tööle pandud ja uuritud. See tõesti on ilmselt paratamatu ja pigem positiivne muutus meditsiini tulevikus. See ilmselt toob kaasa meditsiinis osade erialade tööprotsesside suured muutused (puudutab alguses kõige enam neid, kes tegelevad nö mustrite äratundmisega nagu radioloogid, patoloogid, dermatoloogid). Raamatus toob autor välja ka AI kitsaskohti (kasvõi AI rakenduse tegemisel loodud viga või eelarvamus, andmete kaitsmine, sisendi ebaühtlus või sisestamise vaevanõudmine ja sellest tingitud tulemuse eksimus). Lõpuks jõuab autor arsti-patsiendi suhte ja inimlikkuse olulisuseni, mida AI kunagi ei saa päriselt asendada. Kokkuvõttes põnev sissevaade teistmoodi tulevikumaailma. Ma ise ei suutnud täielikult nautida teost, sest lihtsalt ajumaht on sügisel nii palju muude asjadega tegelenud. Meditsiini ja AI huvilistele soovitan aga küll. Raamat võimaldab kindlasti alustada erinevaid ägedaid arutelusid meditsiinieetika ja AI ja õppe teemadel.
Read
May 12, 2019Terminei de ler o livro Deep Medicine, How Intelligence can make healthcare human again, escrito por Eric Topol. É um livro muito atual, escrito nesse ano e lançado mês passado.
Ele conta no livro a história da evolução da Inteligência Artificial, Machine Learning, CNNs, Self driving vehicles, Speech Recognition NN, Deep Face facial recognition, sobre o desafio vencido por um algoritmo jogando AlphaGo com humanos e o lançamento do Intituto AI Now em 2017. Ele explica como cada uma dessas tecnologias avançaram e utiliza exemplos na Medicina pra ilustrar esse progresso.
Mostra experimentos feitos com reconhecimento de imagens e diversas startups que estão trabalhando para trazer melhores diagnósticos, baseado em aprendizado em imagens de exames médicos. Traz também avanços na psicologia e psiquiatria usando informações de reconhecimento de voz, além do uso de assistentes digitais.
Para alimentação, algumas startups vêm propondo dietas personalizadas baseadas nos nossos dados e material genético.
Mas a Empatia é uma virtude humana que nunca poderá ser criada pela máquina ou algoritmo. O entendimento do sofrimento e sensibilidade a outra pessoa é algo único, característica do ser humano.
Eric ainda aponta uma preocupação muito grande sobre o uso dos nossos dados vitais por grandes empresas, que hoje está de forma desfragmentada, mas que deveria ser nossa propriedade e que nós deveríamos ter acesso a esses dados e, com eles, quando necessário, levá-los a especialistas, empresas, instituições ou hospitais para que pudéssemos melhorar nossa qualidade de vida, nos tratar, ou mesmo salvar nossa vida em caso de doença grave.
É um livro recomendo tanto pros amigos e amigas que trabalham com cuidado a vida, como enfermeiras, médicas, cuidadoras, quanto todas as pessoas que quiserem se atualizar com os avanços da tecnologia aplicados ao nosso dia a dia e convido a pensar sobre as possíveis implicações e impactos dessa tecnologia para a maior longevidade do homem, melhora de sua qualidade de vida e impacto na sociedade.
Pra quem topar, boa leitura:
https://www.amazon.com/Deep-Medicine-...
Ele conta no livro a história da evolução da Inteligência Artificial, Machine Learning, CNNs, Self driving vehicles, Speech Recognition NN, Deep Face facial recognition, sobre o desafio vencido por um algoritmo jogando AlphaGo com humanos e o lançamento do Intituto AI Now em 2017. Ele explica como cada uma dessas tecnologias avançaram e utiliza exemplos na Medicina pra ilustrar esse progresso.
Mostra experimentos feitos com reconhecimento de imagens e diversas startups que estão trabalhando para trazer melhores diagnósticos, baseado em aprendizado em imagens de exames médicos. Traz também avanços na psicologia e psiquiatria usando informações de reconhecimento de voz, além do uso de assistentes digitais.
Para alimentação, algumas startups vêm propondo dietas personalizadas baseadas nos nossos dados e material genético.
Mas a Empatia é uma virtude humana que nunca poderá ser criada pela máquina ou algoritmo. O entendimento do sofrimento e sensibilidade a outra pessoa é algo único, característica do ser humano.
Eric ainda aponta uma preocupação muito grande sobre o uso dos nossos dados vitais por grandes empresas, que hoje está de forma desfragmentada, mas que deveria ser nossa propriedade e que nós deveríamos ter acesso a esses dados e, com eles, quando necessário, levá-los a especialistas, empresas, instituições ou hospitais para que pudéssemos melhorar nossa qualidade de vida, nos tratar, ou mesmo salvar nossa vida em caso de doença grave.
É um livro recomendo tanto pros amigos e amigas que trabalham com cuidado a vida, como enfermeiras, médicas, cuidadoras, quanto todas as pessoas que quiserem se atualizar com os avanços da tecnologia aplicados ao nosso dia a dia e convido a pensar sobre as possíveis implicações e impactos dessa tecnologia para a maior longevidade do homem, melhora de sua qualidade de vida e impacto na sociedade.
Pra quem topar, boa leitura:
https://www.amazon.com/Deep-Medicine-...
August 2, 2021
Deep Medicine - though book emphasis on Health care it also covers adducent areas which can be directly influenced by AI. Author Eric Topol , is a renowned practitioner of medicine who has great knack of stepping out of the circle , viewing the subject from outside and providing unbiased commentary on it. He provides a balanced view of the possibility of AI and its limitations.
The essence of the book is covered in the last chapter. It's where the author brings home the thesis about why humans should be the central theme for AI development. Unlike any other field, AI in medicine/ care presents a unique challenge for the inventor. The placebo effect of medicine is as powerful as real medicine in some cases. His view about how adapting AI will help practitioners to go back to the basics of caring for the patients is a fascinating argument.
Overall it is a very good read.
The essence of the book is covered in the last chapter. It's where the author brings home the thesis about why humans should be the central theme for AI development. Unlike any other field, AI in medicine/ care presents a unique challenge for the inventor. The placebo effect of medicine is as powerful as real medicine in some cases. His view about how adapting AI will help practitioners to go back to the basics of caring for the patients is a fascinating argument.
Overall it is a very good read.
Displaying 1 - 30 of 285 reviews