Week 1: What does all of this mean?

Even before I arrived in New York City, I knew I was in for a roller coaster ride. My clinical mentor, Dr. Daniel Margolis, who worked in the Radiology Department, informed me that I would be helping to making “deep neural networks” for the purpose of identifying prostate tumors in magnetic resonance imaging (MRI) scans. Although I had heard of the term deep learning I did not have an inkling of what it meant. This was until I met Dr. Alex Bratt, a previous radiology resident who worked with Dr. Margolis on developing a deep learning algorithm. Dr. Bratt explained that a part of deep learning is to train the algorithm to identify a specific object in an image by inputting many different images of the same object.   

His program identified the prostate, and its regions, the transitional and peripheral zones in MRI scans.  The transitional zone is near the center of the prostate and the peripheral zone is at edge closest to the rectum. Identifying these regions helps to identify tumors as tumors typically develop in the peripheral zone, and allows doctors to the discuss the mode of access for removing the tumor. One of the really amazing things that Dr. Bratt showed me was an image of the prostate tumor that had invaded into bladder, which the computer was able to detect, indicating its robustness to identify prostate tissue in non-prostate regions.

One of the interesting topics of our conversation was how deep learning will affect the future of medicine and whether it has the possibility of putting radiologists out of work. Dr. Bratt’s short answer to this idea was that although deep learning has the possibility of being able to read radiology images perfectly some day, if such complex algorithms have been developed than radiologists won’t be the only ones out of their jobs. However, he did mention that it might not replace radiologists but just make their efficient, so that they are able to read more radiology scans and help more patients.

Later in the week, I was able to shadow Dr. Margolis as he looked over MRI and computed tomography (CT) scans. Although the work was difficult, it was very interesting to see how a small discoloration in an image, practically invisible to the untrained eye, can identify the presence of a tumor, or a blockage, or abnormal tissue. 

On a rather fun note, I visited Central Park for the first, listened to the astounding New York Philharmonics and saw some amazing fireworks. Below is a picture from my visit. Overall, I am very excited to be participating in this summer immersion and I feel as though it will be one of the unforgettable moments of my Ph.D. time.