Pancreatic cancer is a highly lethal and incurable disease of which survival rates have not improved significantly in the past 40 years. Currently, MRI-guided radiotherapy utilizing high-energy ion beams is used to help shrink cancerous tumors. However, due to inter-patient variability, where the manifestation of the pancreas is different in every patient, and infraction anatomical changes such as breathing, tumors do not get treated effectively. Despite the benefits of being non-invasive, radiotherapy tools are not becoming popular for pancreatic cancer due to manual intervention, human error, and movement of the pancreas during treatment. Today, pancreatic cancer oncologists are often forced to overshoot on the targeted pancreas to attempt controlling tumor size. My project proposes a novel artificial-intelligence-based tool to automatically segment out the pancreas in real-time.

With a balance between accuracy and efficiency, Pancreatic Cancer Deep Learning System (PCDLS) is able to segment out the pancreas in real-time from MRI images while a patient is getting treated. It reduces the radiation overlay to 4mm which can save millions of healthy cells, improve patient quality of care, and help cure this deadly disease. I trained PCDLS on a radiologist confirmed dataset of abdominal images. To confirm my research, I contacted over 253 doctors from leading institutions who were able to validate my problem statement and provide a “voice from the customer”. For future work, I want to couple PCDLS with an MRI machine to create an end-to-end system, create population-based screening protocols, and apply for FDA approval of the process.

What inspired you (or your team)?

I have been deeply moved by family friends passing away from cancer, leading me to volunteer at the Fall 2018 Purple Stride 5K run. I met several pancreatic cancer patients and survivors. It was an inspiring moment to hear their personal stories of helplessness with a deep desire to be happy. Surprising statistics like the low survival-rate and aggressiveness hooked me in. As a technology enthusiast, I was inspired by Steve Jobs and knew how he had passed away from pancreatic cancer. I started thinking about what I could do to address the pandemic. While visiting a laboratory in Boston, I learned about pancreatic cancer radiotherapy and medical imaging. I questioned if I could use my knowledge in programming and artificial intelligence to help tackle the pressing problem. Thus, my project was born.