top of page
  • Writer's picturebxgenetics

AI and Cancer Diagnoses

By: Arham Akhyer

AI has seen a surge in popularity in the recent few years. One of the largest sectors that AI has caught the eye of many people is in medicine. AI has been researched and studied especially when it comes to fatal diseases. One of these main diseases is cancer which takes the lives of 600,000 Americans each year. Fortunately, AI has shown some promising results that may help in the detection of the early onset of the disease.

DeepGlioma is an AI-based diagnostic screening system developed by neurosurgeons and engineers at the University of Michigan. This system uses swift imaging of tumor specimens collected during operations. These specimens are then analyzed for genetic mutations.

In a study involving over 150 patients with glioma, DeepGlioma had correctly associated mutations with specific molecular subgroups more than 90% of the time. Diagnosis and treatment of glioma is premised on the molecular conditions detected and has high variance from patient to patient, which is why DeepGlioma is such a revolutionary breakthrough. Before DeepGlioma, doctors had a very difficult time trying to differentiate glioma from each other. Plus, it would typically take much longer than 90 seconds for humans to do this process via the traditional route using optical imaging of neural networks which just shows how efficient AI in medicine can be when developed correctly.

Another AI tool named Sybil is being tested and developed. Scientists from the Mass General Cancer Center and MIT created Sybil to predict the onset of lung cancer. One study has shown that Sybil accurately predicted if a person would develop lung cancer in a year 86% to 94% of the time. Sybil allows scientists to see things that would be invisible to radiologists via CT scans. In fact, sometimes Sybil allows doctors to see development years before they would have otherwise. From a single CT Scan, Sybil may be able to save a person’s life since early detection of lung cancer is a crucial confounding factor influencing survival chances.

Of course, Sybil and DeepGlioma are not the only AI tools being used in medicine. Over 300 AI tools have been FDA approved in radiology alone. However, this popularity also has led to the concern with the data that is being used to train these AI. Much of the data used does not come from diverse communities within the United States and is not representative of the entire nation. Thus, many scientists are arguing that AI tools are not being developed equally and as efficiently for black and brown communities.

Regardless, since AI is such a novel topic of discussion there is no doubt much more research needed and hours put in to effectively normalize AI in clinical settings for everyone.


10 views0 comments


bottom of page