Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) are combining big data sets from previous cancer diagnoses and new algorithms to more accurately diagnose cancer.
Yuan Luo, a PhD student in Electrical Engineering and Computer Science (EECS), and EECS Professor Peter Szolovits have developed a computer model that can help doctors more easily and more accurately distinguish between forms of lymphoma. The pair is working with a team from Massachusetts General Hospital.
Just as cancer itself is not one disease, lymphoma—a cancer of the lymphatic system—has about 50 distinct subtypes. Between 5–15% of lymphatic cancer cases are misdiagnosed or misclassified, which can result in patients receiving the wrong treatment plan.
Luo’s computational model analyzes large data sets from thousands of cancer patients and recommends a specific diagnosis of the disease. As Luo explained in a recent article in Software Development Times, the task of diagnosing these cancers is time-consuming and labor intensive. “The amount of patient data exceeds the capacity of human experts,” he said.
While analyzing data is a critical step, Luo’s model goes further, providing interpretation of the data. “Often software developers build a machine-learning pipeline that has pretty good accuracy, but you hear domain experts (in our case, pathologists) complaining that they don’t understand what the model is doing. Our method tries to bridge the interpretability gap. It can automatically capture relations between concepts.”
Professor Szolovits hopes that this new model can not only make lymphoma diagnosis more accurate, resulting in better patient care and outcomes, but also that the information can be incorporated into future World Health Organization guidelines, which to date have been based on significantly smaller samples of the subtypes.