Skip to content 
Massachusetts Institute of Technology( MIT) Dr. Regina Barzilay has been recognised in TIME Magazine’s 2025 list of artificial intelligence leaders, innovators, shapers and thinkers, called TIME100 AI 2025. Barzilay has worked on issues like spearheading the creation of Mirai, an AI that uses mammograms to predict the likelihood of breast cancer, and Sybil, an AI that uses CT scans to estimate the risk of lung cancer. Both tools have already been widely tested and validated across hospitals around the world and are currently being prepared to be integrated into clinical settings. Her identification emphasises the way AI is transforming cancer detection and survival rates.
Dr. Regina Barzilay's Diagnosis That Sparked her Innovation
Barzilay started her research when she was diagnosed with breast cancer in 2014. She remembered that she had no family history of BRCA mutations and had a healthy lifestyle, but still had to deal with the disease. The experience motivated her to develop solutions that would help to identify cancer earlier and more precisely. In September 2019, the AI model Mirai, created with MIT students, was released. It uses mammograms and estimates the risk of breast cancer development within five years. Barzilay states that the tool would have identified her as being at high risk two years before her doctors did.
She described that on closer viewing of a mammogram she had in 2012 and 2013, radiologists missed some indication of disease that Mirai would have identified. The World Health Organisation suggests that breast cancer is the leading cancer in women throughout the world. Half of the cases occur without apparent risk factors other than age and gender. Barzilay has observed that most of the women who are diagnosed very late are those who go to get a regular mammogram. The early diagnosis not only enhances the survival rates but it also minimises the use of vigorous therapy, like chemotherapy and hair loss-inducing drugs.
Global Validation of Mirai
Mirai has been tested on over two million mammograms in 70 hospitals in 22 different countries. These mass studies show that it can be applied in different populations. Barzilay pointed out that the model demonstrates how AI can identify trends that cannot be seen by people.
It is an aid that is supposed to assist the doctors by giving them probability-based forecasts and not to take up clinical judgment. Mirai presents quantified risks, which give patients and physicians a more solid basis on which to make decisions about treatment.
Lung Cancer Prediction with Sybil
In addition to Mirai, Barzilay helped create Sybil, an AI system that predicts six-year lung cancer risk using low-dose CT scans. The project was developed in collaboration with Massachusetts General Hospital and was inspired by a personal connection. A friend of Barzilay’s sister, who had never smoked and appeared healthy, developed a persistent cough. Her doctor initially did not consider lung cancer, leading to delays in testing. By the time an X-ray was performed, the disease had advanced, and she died in her 50s.
Barzilay stated that Sybil was tested across patients of different ages, genders, and races to ensure accuracy. Lung cancer remains the deadliest cancer worldwide, with the Centres for Disease Control and Prevention reporting 131,584 U.S. deaths in 2023 alone. To expand its reach, Sybil is undergoing regulatory review by the Taiwanese Food and Drug Administration, a move prompted by high lung cancer rates in Taiwan and other parts of Asia.
Barzilay AI in Drug Discovery Beyond Cancer
Barzilay has also applied AI to pharmaceutical research. Working with fellow MIT professor Jim Collins, she co-led the creation of a model that screened over 100 million molecules, resulting in the discovery of the first antibiotic identified by AI in 2020. This work illustrates AI’s broader potential in reducing trial-and-error approaches within medicine.
Her overarching focus has been on addressing uncertainty in healthcare. She noted that medical predictions often involve probabilities rather than certainties, leaving patients and doctors without precise answers. AI models like Mirai and Sybil aim to reduce this uncertainty by processing vast amounts of data and generating likelihoods based on patterns.
Addressing Trust in AI for Healthcare
Barzilay acknowledged concerns about artificial intelligence in medicine, particularly as public attention has centred on tools such as ChatGPT. She stressed that clinical AI tools are narrowly focused, thoroughly validated, and must be FDA-approved before use. She compared AI adoption in medicine to the unquestioned trust patients place in routine processes like blood testing, where safeguards ensure accuracy.
Barzilay said that once AI tools undergo rigorous validation and regulatory review, they should be regarded with the same confidence as established diagnostic methods. Mirai and Sybil have already been validated in hospitals across multiple countries, strengthening their credibility. However, Barzilay identified a remaining challenge: establishing clinical protocols for patients identified as high risk. Predictive tools must not only detect disease but also guide follow-up actions to improve outcomes.Her recognition in TIME100 AI underscores the progress of these technologies but also the work still ahead. Barzilay stated that the ultimate goal is not just prediction, but meaningful improvement in patient outcomes.
