Deep Learning for Cancer Diagnosis and Treatment: A Personal and Scientific Journey with Regina Barzilay

Regina Barzilay, an MIT professor specializing in Natural Language Processing and deep learning, discusses the transformative potential of AI in oncology, particularly for early cancer diagnosis, prevention, and treatment. She highlights the profound impact of non-technical literature, citing "The Emperor of All Maladies" for its insights into the imperfect, human-driven nature of scientific discovery, and "Americana" for its reflections on cultural adaptation. A central argument is that while scientific ideas are crucial, the devotion and personality of individuals significantly accelerate the adoption and implementation of advancements, a factor often overlooked in the purely objective pursuit of science.

Barzilay draws a critical distinction between the traditional emphasis on mechanistic understanding in biology and the probabilistic matching approach prevalent in computer science. She posits that given the immense complexity of the human body, a full deterministic understanding might be beyond our current capacity, suggesting that AI's strength lies in pattern recognition and prediction without necessarily achieving deep, human-like comprehension. Her personal battle with breast cancer in 2014 profoundly shifted her perspective, redirecting her focus from what she perceived as "trivial" academic pursuits to applying her expertise to alleviate real-world suffering, particularly in healthcare.

The podcast emphasizes the critical role of machine learning in early cancer detection, which is often a life-saving factor for diseases like pancreatic cancer, where late diagnosis is typically a death sentence. Barzilay points out severe challenges in data access, noting the lack of publicly available modern medical datasets (e.g., mammograms) and the legal and incentive barriers hospitals face in sharing data. She advocates for patient-driven data donation mechanisms, similar to organ donation, to overcome these hurdles. Technical solutions like de-identification and learning on encoded data are proposed to address privacy concerns, alongside public education about the societal benefits of data sharing.

The conversation broadens to the wider implications of AI in healthcare, including accelerating drug discovery and addressing other severe conditions like neurodegenerative diseases. Barzilay expresses concern about the slow pace of adoption by the medical establishment and regulatory bodies (such as the FDA), which she identifies as a major bottleneck. The historical failures of initiatives like Google Health and Microsoft HealthVault underscore the complex interplay of regulatory, business, and cultural factors that hinder large-scale data collection and innovation in healthcare, suggesting a need for a more holistic, perhaps "anthropological," understanding of these systemic challenges.

that book despite the fact that I am in the business of science really opened my eyes on how imprecise and imperfect the discovery process is and how imperfect our current solutions

it's not necessarily that they are more important than ideas but I think that ideas on their own unknown sufficient and many times at least at the local horizon is the personalities and their devotion to their ideas is really the locally changes the landscape

if you're looking about recommendation system they're not claiming they understanding somebody they're just managing to from the patterns of your behavior to recommend you a product

I look back at what my group was doing what other groups was doing and I saw these trivialities it's like people are building their careers on improving some parts around two or three percent or whatever I was like seriously

why are we trying to improve the parser or deal with some trivialities when we have capacity to really make a change

we all need to join the battle [against diseases like cancer and neurodegenerative diseases]

what machine learning can do here is utilize all this data to tell us le who is like it'll be susceptible and using all the information that is already there beat imaging beat your other tests and you know eventually liquid biopsies and others where the signal itself is not sufficiently strong for human eye to do good discrimination

like right now in this country there is no publicly available data set of modern mammograms that you can just go on your computer sign a document and get it it just doesn't exist

at that moment you feel that your life is at the stake but you just don't have information to make the choice

it's sad because unfortunately and I have I need to research why that happened but I'm pretty sure Google Health and Microsoft's HealthVault or whatever it's called both closed down which means that there was either regulatory pressure or there's not a business case or there's challenges from hospitals which is very disappointing