MA Researchers Create AI Algorithm to Enhance Lung Cancer Radiation Therapy

Researchers from the Artificial Intelligence (AI) in Medicine Program of Brigham and Women’s Hospital created a deep-learning algorithm to improve lung cancer radiation therapy treatment and provider practices.

Cancer incidence appears to be on the rise, with lung cancer being the most common cancer in the world. This trend is accompanied by a gradual decline in radiation and oncology specialists, leading to strains in healthcare delivery, according to the press release.

To advance lung cancer treatment, Brigham and Women’s Hospital researchers have created an AI model to detect non-small cell lung cancer (NSCLC) tumors within computed tomography (CT) scans.

To train the model to differentiate between tumors and other tissues, researchers used CT images from 787 patients. They tested the model's performance using scans from more than 1,300 patients.

Then, researchers asked eight radiation oncologists to perform segmentation processes and rate and edit segmentations produced by another physician or the AI model.

Researchers found no major difference in performance between segmentations produced by human-AI collaborations and human-only segmentations.

The goal of the project was not only to enhance the treatment for lung cancer patterns but also to increase knowledge on applying AI to clinical care practices.

“The biggest translation gap in AI applications to medicine is the failure to study how to use AI to improve human clinicians, and vice versa,” said corresponding author Raymond Mak, MD, of the Brigham’s Department of Radiation Oncology, in a press release.

The research also shows that radiation oncologists edited segmentations at a pace that was 65 percent faster for those produced by human-AI collaboration compared to those made by a human alone, though they were not told beforehand which was which. Further, the radiation oncologists rated the quality of AI-produced segmentations more highly than the human-produced segmentations.

“We’re studying how to make human-AI partnerships and collaborations that result in better outcomes for patients. The benefits of this approach for patients include greater consistency in segmenting tumors and accelerated times to treatment. The clinician benefits include a reduction in mundane but difficult computer work, which can reduce burnout and increase the time they can spend with patients,” said Mak.

Practices that include the use of an AI algorithm to track and treat conditions have exploded in recent years.

In August, researchers from the University of Florida created an AI algorithm to track COVID-19 variants. They trained the algorithm using known genetic sequences of the coronavirus, allowing it to detect unfamiliar strains and whether or not they attack cells.

In May, Mayo Clinic created an AI algorithm to enhance the process of detecting weak heart function.

And similarly, in April, researchers from the RIKEN Center for Advanced Intelligence Project implemented an AI model to enhance the detection of heart disease in ultrasounds. Following its implementation, they found that diagnostic accuracy improved.