2020 BSJ Blog Fall 2020

Artificial Intelligence: A Potential Platform for Breast Cancer Diagnosis

By Liane Albarghouthi

When you hear of artificial intelligence, what first comes to mind? You may not immediately jump to breast cancer screenings — but with the developments of exciting technologies, this seemingly unlikely pairing is within reach. Now, scientists are unraveling trailblazing artificial intelligence applications throughout a vast range of fields — including cancer diagnosis.  

One of the leading causes of death from cancer for women, breast cancer is typically diagnosed after conducting screening tests, notably the mammogram: an X-ray of the breast. Yet radiologists often experience difficulty interpreting them — leading to inaccurate results. Too frequently, expert clinicians diagnose patients with false positives or false negatives, which undoubtedly unlocks a Pandora’s box of medical and emotional turmoils.

To curtail these problems, researchers have developed an artificial intelligence (AI) system capable of detecting breast cancer with greater accuracy than even the best radiologists. Artificial intelligence — a branch of computer science concerned with developing machines and applications to perform tasks as analytical tools— is used to scan and interpret mammograms for possible signs of breast cancer. 

The system is composed of three deep learning models, each focusing on varying levels of analysis of the breast tissue. Each sub-model then reports a cancer risk outcome between 0 and 1 — with the final diagnosis being an average of the three scores. As such, the AI system detects hidden tumors concealed by the breast tissue with sensitivity and specificity exceeding that of radiologists.

 To assess the potential of this application, Scott Mayer McKinney et al first trained the AI system using a staggering 25,856 mammograms from the United Kingdom and 3,097 mammograms from the United States. The mammograms of biopsy-confirmed breast cancer patients were subsequently used to evaluate the AI system’s precision in diagnosing positives. Lastly, a clinical research organization conducted a study to compare the system’s predictions with that of human experts. Six expert radiologists in the United States analyzed a random sample of 500 mammograms. Using the same sample, the AI system’s predictions were modeled alongside the radiologists’ results. The research team found that the AI model outperformed expert radiologists in accurately diagnosing breast cancer based on the mammograms it scanned. This outcome represents an exciting leap for diagnostic medicine. 

Of course, this form of cutting-edge technology must undergo further trials before it can break the surface and emanate in the real world. For starters, the authors expressed valid concern that the U.S. data pool was not reflective of the breast cancer screening demographics across the country in reference to the fact that all 3,097 mammograms came from just one screening center. Perhaps an even greater indicator for the necessity of further trials is the aspect that most images used in the research were curated from devices manufactured by a singular medical imaging company. To better assess the expediency of this AI system, future research should include images from other brands of medical equipment. 

Based on this study, we can see how AI offers exciting, revolutionary ideas and applications in healthcare today — but we mustn’t jump to conclusions just yet. While there are many barriers to break down before such a system can be implemented in medicine, who knows? 

A greater understanding of AI may potentially set in motion a future of medicine in which it plays a momentous role.

References:

McKinney, S., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., . . . Shetty, S. (2020, January 01). International evaluation of an AI system for breast cancer screening. Retrieved September 26, 2020, from https://www.nature.com/articles/s41586-019-1799-6

Pisano, E. (2020, January 01). AI shows promise for breast cancer screening. Retrieved September 26, 2020, from https://www.nature.com/articles/d41586-019-03822-8

What Is Breast Cancer Screening? (2020, September 14). Retrieved October 1, 2020, from https://www.cdc.gov/cancer/breast/basic_info/screening.htm

What Is a Mammogram? (2020, September 14). Retrieved October, 2020, from https://www.cdc.gov/cancer/breast/basic_info/mammograms.htm

Serokell. (2020, November 23). Artificial Intelligence vs. Machine Learning vs. Deep Learning: What’s the Difference. Retrieved November, 2020, from https://medium.com/ai-in-plain-english/artificial-intelligence-vs-machine-learning-vs-deep-learning-whats-the-difference-dccce18efe7f