Verified by 
Artificial Intelligence
AI Helps Reduce Excessive Radiation for Cancer Patients
AI spots patients with HPV cancer who need less intensive radiation therapy.
Posted January 11, 2022 Reviewed by Ekua Hagan
Key points
- Human papillomavirus (HPV) is attributed to 70 percent of cancer located in the mouth or back of the throat.
- Radiation therapy is often used to treat cancer, but it can also kill healthy cells, impacting the quality of life for patients.
- In a recent study, scientists used AI to identify cancer patients whose treatment could be decreased without impacting the cure rates.
A new study in the Journal of the National Cancer Institute shows how artificial intelligence (AI) machine learning can identify HPV-attributed oropharyngeal cancer patients who require less aggressive radiation in order to achieve a favorable outcome.
Human papillomavirus (HPV) is attributed to 70 percent of cancer located in the mouth or back of the throat (oropharyngeal cancer or oropharyngeal squamous cell carcinoma) according to the U.S. Centers for Disease Control and Prevention (CDC). In 2021, there were over 54,000 new cases and 10,850 died of oral cavity and oropharyngeal cancers in the United States according to estimates from the American Cancer Society.
Radiation therapy is often used to treat cancer either alone or in combination with surgery, chemotherapy, or other treatments according to the American Cancer Society. In radiation therapy, high doses of strong beams of energy (protons, electron beams, X-rays, or gamma rays) are used to kill the tumor. Sometimes, healthy cells are killed in the process which can impact the quality of life for cancer patients.
Identifying low-risk HPV cancer patients using OP-TIL
The study was led by Anant Madabhushi, the Donnell Institute Professor of Biomedical Engineering at Case Western Reserve University and head of the Center for Computational Imaging and Personalized Diagnostics (CCIPD), and Germán Corredor Prada, a research associate in the CCIPD lab.
The scientists sought to use AI to identify cancer patients who are not only at minimal risk of recurrence and death but also whose treatment could be decreased without impacting the cure rates. To achieve this goal, the team developed an imaging biomarker called OP-TIL that could identify which could separate cancer patients into categories of either high or low-risk groups in order to identify candidates for de-escalation clinical trials.
OP-TIL is aptly named, as a biomarker for tumor-infiltrating lymphocytes (TILs). Lymphocytes are a type of white blood cell that protects the body from infections. The three main types of lymphocytes are natural killer (NK) cells, T cells, and B cells. Tumor-infiltrating lymphocytes are white blood cells that have left the bloodstream towards a tumor.
According to the researchers, TILs “appear to have a protective effect through an adaptive host immune response directed against viral antigens” and “an increased density of TILs is associated with low risk of recurrence in low-stage HPV-associated OPSCC.”
“OP-TIL represents a potentially powerful, low-cost, and easy-to-scale imaging-based biomarker that may be able to select out the patients from the so-called low-risk HPV-associated OPSCC cohort who are nevertheless destined to recur,” reported the researchers. “This can spare them from ill-advised de-escalated therapy and thereby enhance the outcomes for the remaining truly low-risk patients for whom de-escalation is more appropriate.”
To create the imaging biomarker Op-TIL, the researchers used actual samples from 94 cancer patients for model training and feature discovery and assessed it on samples from 345 patients.
The team used an unmodified trained deep learning model, specifically a generative adversarial network (GAN), that was created in a separate study by Faisal Mahmood at Brigham and Women’s Hospital, Harvard Medical School, and the Broad Institute of Harvard and MIT along with his research colleagues that was published in IEEE.
The AI algorithm uses image data as input and outputs the location of lymphocytes and non-lymphocytes within the image. The lymphocyte algorithm was trained on lung images. Despite the fact that lymphocytes do not vary in appearance much across various organs, the team validated the quality of the output with an expert human pathologist.
The researchers reported that their solution was able to identify cancer patients who would have benefited from greatly reduced doses of radiation therapy.
“OP-TIL can identify stage I HPV-associated OPSCC patients likely to be poor candidates for treatment de-escalation,” the researchers wrote. “Following validation on previously completed multi-institutional clinical trials, OP-TIL has the potential to be a biomarker, beyond clinical stage and HPV status, that can be used clinically to optimize patient selection for de-escalation.”
Copyright © 2022 Cami Rosso All rights reserved.
