On 27 March 2024, Professor Regina Barzilay, faculty lead for artificial intelligence (AI) at the MIT Jameel Clinic, appeared on the PBS Nova documentary 'AI revolution' to discuss the role of AI in cancer detection. An excerpt from the film and transcript are below. The full documentary can be watched here: https://www.pbs.org/wgbh/nova/video/ai-revolution/
Miles O'Brien: Computer scientist Regina Barzilay first started working on artificial intelligence in the 1990s, just as rule-based AI, like Deep Blue, was giving way to neural networks. She used the techniques to decipher dead languages. You might call it a small-language model.
Regina Barzilay: Something that is fun and intellectually very challenging, but it’s not like it’s going to change our life.
Miles O'Brien: And then, her life changed, in an instant.
Constance Lehman: We see a spot there.
Miles O'Brien: In 2014, she was diagnosed with breast cancer.
Regina Barzilay: When you go through the treatment, there are a lot of people who are suffering. I was interested in what I can do about it, and, clearly, it was not continuing deciphering dead languages. And it was quite a journey.
Miles O'Brien: Not surprisingly, she began that journey with mammograms.
Constance Lehman: It’s a little bit more prominent.
Miles O'Brien: She and Constance Lehman, a radiologist at Massachusetts General Hospital, realised the Achilles heel in the diagnostic system is the human eye.
Regina Barzilay: So, the question that we ask is, what is the likelihood of these patients to develop cancer within the next five years? We, with our human eyes, cannot really make these assertions because the patterns are so subtle.
Constance Lehman: Now, is that different from the surrounding tissue?
Miles O'Brien: It’s a perfect use case for pattern recognition, using what is known as a convolutional neural network.
Here’s an example of how CNNs get smart: they comb through a picture with many virtual magnifying glasses. Each one is looking for a specific kind of puzzle piece, like an edge, a shape or a texture. Then it makes simplified versions, repeating the process on larger and larger sections.
Eventually the puzzle can be assembled. And it’s time to make a guess. Is it a cat? A dog? A tree? Sometimes the guess is right, but sometimes it is wrong. And here’s the learning part: with a process called backpropagation, labelled images are sent back to correct the previous operation. So, the next time it plays this guessing game, it will be even better.
To validate the model, Regina and her team gathered up more than 128,000 mammograms collected at seven sites in four countries. More than 3,800 of them led to a cancer diagnosis within five years.
Regina Barzilay: You just give to it the image and then the five years of outcomes. And it can learn the likelihood of getting a cancer diagnosis.
Miles O'Brien: The software, called Mirai, was a success. In fact, it is between 75 and 84 percent accurate in predicting future cancer diagnoses. Then, a friend of Regina’s developed lung cancer.
Lecia Sequist: In lung cancer, it’s actually sort of mind boggling how much has changed.
Miles O'Brien: Her friend saw oncologist Lecia Sequest. She and Regina wondered if artificial intelligence could be applied to CT scans of patients’ lungs.
Lecia Sequist: We taught the model to recognise the patterns of developing lung cancer by using thousands of CT scans from patients who were participating in a clinical trial. We had a lot of information about them. We had demographic information, we had health information, and we had outcomes information.
Miles O'Brien: They call the model Sybil.
Florian Fintelmann: In the retrospective study, the retrospective data–
Miles O'Brien: Radiologist Florian Fintelmann showed me what it can do.
Florian Fintelmann: This is earlier, and this is later. There is nothing that I can perceive, pick up or describe. There’s no, what we call a precursor lesion on this CT scan. Sybil looked here and anticipated that there would be a problem, based on the baseline scan.
Miles O'Brien: What is it seeing?
Florian Fintelmann: That’s the million dollar question and, and maybe not the million dollar question. Does it really matter? Does it?
Miles O'Brien: When they compared the predictions to actual outcomes from previous cases, Sybil fared well. It correctly forecast cancer between 80 and 95 percent of the time, depending on the population it studied. The technique is still in the trial phase. But once it is deployed, it could provide a potent tool for prevention.
Regina Barzilay: The hope is that if you can predict, very early on, that the patient is in the wrong way, you can do clinical trials, you can develop the drugs that are doing the prevention, rather than treatment of very advanced disease that we are doing today.