Artificial intelligence can augment existing methods for predicting the risk of head and neck cancer spreading beyond the boundaries of neck lymph nodes, according to researchers at the ECOG-ACRIN Cancer Research Group (ECOG-ACRIN). A custom deep learning algorithm using standardized CT scan images and associated data contributed by patients who participated in the E3311 phase 2 trial shows promise, especially for patients with a new diagnosis of HPV associated with head and neck cancer. The E3311 validated dataset holds the potential to contribute to more accurate disease staging and risk prediction.
Benjamin Kahn, (MD) (Dana-Farber Cancer Institute, Brigham and Women’s Hospital, Harvard Medical School), led the ECOG-ACRIN study. He will present the findings during the annual meeting of the American Society of Radiation Oncology (ASTRO) in San Antonio, Texas.
“This type of research is essential because it can help identify patients with high-risk and aggressive diseases and also help select suitable patients for de-escalation therapy,” said Dr. Kahn.
Head and neck cancers and their standard treatments – surgery, radiotherapy or chemotherapy – carry significant morbidity. They affect a person’s appearance, speech, eating or breathing. Therefore, there is great interest in developing less intensive treatment strategies for patients. For example, the completed Phase III E3311 trial showed that low-dose radiation at 50 Gray (Gy) without chemotherapy after oral surgery resulted in very high survival and outstanding quality of life in patients at average risk of recurrence (RL phrases. J Clin Unk. December 2021).
Dr. Kahn and colleagues developed and validated a neural network-based deep learning algorithm based on diagnostic CT scans, pathology, and clinical data. The source was the group of participants in the E3311 trial who were assessed as being at risk of recurrence by standard pathological and clinical measures.
“The staging of head and neck cancers is a challenging clinical problem,” said Dr. Kahn. “In particular, our current efforts to quantify external extension by human interpretation of pre-treatment imaging have generally shown poor results.”
Factors determining the stage of cancer include the size of the original tumor, the number of lymph nodes involved, and external extension – when malignant cells spread beyond the boundaries of the lymph nodes in the neck to surrounding tissues. At E3311, patients were assessed as high risk if there was a 1 mm external extension (ENE). These patients were assigned to chemotherapy and high-dose radiation (66 Gy) after transoral surgery.
Dr. Kahn and colleagues obtained pre-treatment CT scans and corresponding surgical pathology reports from the E3311 high-risk group, as available. Of the 177 scans collected, 311 nodes were annotated: 71 (23%) with ENE and 39 (13%) with ≥1 mm ENE.
The tool showed a high performance in the ENE prediction, significantly outperforming reviews by head and neck radiology experts.
“The deep learning algorithm rated 85% of nodes as having ENE compared to 70% by radiologists,” Dr. Kahn said. “In terms of specificity and sensitivity, the deep learning algorithm was 78% accurate versus 62% by radiologists.”
The team plans to evaluate the data set as part of future treatment trials for head and neck cancer. The algorithm will be evaluated for its ability to improve current disease risk assessment and staging methods.
“Our ability to develop biomarkers from standard CT scans is an exciting new area of clinical research and offers hope that we will be able to better tailor treatment to individual patients, including determining when surgery is best used and who should minimize the extent of treatment,” Kabir said. Authors Barbara A. Bertens, MD.
Dr. Bertens is Professor of Medicine and co-leader of the Developmental Therapy Research Program at Yale Cancer Center, chair of the ECOG-ACRIN Head and Neck Committee, and chair of the ECOG-ACRIN Task Force on the Advancement of Women.
Summary 141 (External extension check with deep learning: Evaluation in ECOG-ACRIN E3311, a randomized de-escalation trial of HPV-associated oropharyngeal carcinoma. is a science event at the “Subject Variations: Strategies for HPV + Oropharyngeal Cancer Decondensation” on Monday, October 24 at 11:15 AM CST.
ASTRO awarded Dr. Kahn a Basic/Transitional Science Award for his exploration of the novel.
Dr. Kahn will also be one of the committee members in educational session Designed for Head and Neck Cancer Practitioners on Wednesday, October 26th from 8:00 – 9:00 AM Central Time. The session aims to break down barriers to understanding AI and promote adoption in the future.
for more information, View Dr. Kahn’s profile on the ASTRO website.
Co-authors include Benjamin H. Kahn, Jirapat Leketlerswang, Zizong Yi, Sanjay Anega, Henry S. Park, Richard Bakst, Hilary R. Kelly, Amy F. Giuliano, Sam Payapvash, Jeffrey B. Subramaniam, Robert L Ferris and Barbara A Bertens
This study was supported by the ECOG-ACRIN Cancer Research Group (Peter J. Award numbers: U10CA180794, U10CA180820, UG1CA233180, UG1CA233184, UG1CA233337, UG1CA233253, and UG1CA232760.
Disclaimer: AAAS and EurekAlert! is not responsible for the accuracy of newsletters sent on EurekAlert! Through the contributing institutions or for the use of any information through the EurekAlert system.