On Sep. 24, a team of New York University researchers published their findings on using artificial intelligence for breast cancer detection in the scientific journal Nature. One of the co-leaders of this project is Farah Shamout, Assistant Professor Emerging Scholar of Computer Engineering at NYU Abu Dhabi.

Compared to traditional methods such as mammography and human examination by radiologists, the AI system drastically decreases the rate of false positives by almost 40 percent. In addition, it also reduces the number of unnecessary biopsies, which involve taking a small sample of body tissue or cells for further analysis.

While mammography is the current standard for breast cancer detection, it is less sensitive when it comes to women with dense breast tissue and is generally very expensive. Shamout explained that in women with dense breast tissue, ultrasound imaging is recommended as a supplementary modality, which is where the AI comes into play.

The AI system analyzed more than five million ultrasound images and produced a model which scored higher in accuracy in comparison to ten board certified breast radiologists. However, this isn’t to say that the value of human examination is entirely undermined, or that AI is here to replace medical professionals.

“The collaboration between the AI and the human radiologist actually achieved the best performance, so this shows that in the future, women may benefit significantly from the sort of human-AI collaboration in cancer detection, hence [improving] patient outcomes eventually,” Shamout summarized the significance of her study.

Based primarily at the NYU Grossman School of Medicine, Shamout worked on this project with other NYU New York researchers during her time as a visiting scholar there. When the Covid-19 pandemic hit in March 2020, she had to transition to working remotely as she moved back to Abu Dhabi. Despite this, Shamout and her team have made tremendous progress in their research. She envisions that their findings will translate into real world technologies in the next five years and hopefully will have an impact on public policy as well.

“Screening programs currently recommend ages 40 above or 50 and above, depending on [national guidelines], but for women with dense breast tissue it might be preferable to start with a different imaging modality first,” she pointed out. “[The AI system can] potentially impact policy and how screening programs are basically designed for women.”

Going forward, the team is looking to improve the system’s robustness and accuracy. They are interested in seeing how the AI would perform across different populations, for example in the UAE, as most of their existing data has been in the US so far. They are also looking to combine different data modalities such as mammography and ultrasound imaging as they contain complementary information.

Keenly interested in engineering applications and healthcare, the young professor did her PhD on AI in healthcare at the University of Oxford. Before that, she completed her undergraduate education here at NYUAD, graduating in 2016 with a degree in computer engineering.

During her time at NYUAD, Shamout became more aware of and interested in medical issues pertinent to the region, giving her insight into problems that could use the help of AI applications. The relevant research assistantships that she did as an undergraduate also solidified her interest in the field and encouraged her to pursue her PhD.

To students who are interested in learning more about the intersection between AI and healthcare, Shamout encouraged them to get informed and get involved: “There are a lot of resources out there now, so many papers and online courses … Expand [your] knowledge on a specific area and consider doing a research assistantship with a faculty member that works in the field.”