AI-ing Microenvironment Improves Prediction of Extracapsular Nodal Extension in Oropharyngeal Carcinoma

Purpose/Objective(s)Deintensification for early-stage oropharyngeal carcinoma includes surgery +/- adjuvant radiation therapy without the use of concurrent chemotherapy. Pathologic findings of nodal extracapsular extension (ECE), however, obligate the use of postoperative concurrent chemotherapy in these patients. Machine learning (ML) and deep learning (DL) algorithms using clinical annotations or CT images have yielded less than optimal ECE prediction with AUCs of 0.58-0.85. We hypothesized that interplay between involved nodes and peritumoral environment plays a pivotal role in ECE determinants. The purpose of this study was to investigate if AI-ing the microenvironment of nodal metastasis can enhance ECE prediction.Materials/MethodsBetween 2016 and 2022, 171 patients with newly diagnosed resectable oropharyngeal carcinoma underwent upfront transoral robotic surgery and neck dissection at our institutions. In all these patients, high-resolution CT scans were done preoperatively and were concluded to have no evidence of ECE by experienced head and neck radiologists. Median age was 63 years. Radiomics and topological features from the nodal and perinodal microenvironment were extracted to generate predictive models.ResultsThere was a total of 264 pathological lymph nodes. Of these positive nodes, 20% exhibited ECE with a ECE of <1 to 10mm. Extraction of radiomics features of the nodal microenvironment, in addition to nodal metastasis, improves AUC from 0.76 to 0.87. Analysis of topological features of the perinodal region, in addition to nodal metastasis, improves AUC from 0.61 to 0.79. Multimodal predictive model employing both radiomics and topological features yields an AUC of 0.94. Inclusion of HPV status in our multimodal further improved the performance with an AUC of 0.95, a true-positive rate of 100%, and a true-negative rate of 90%.ConclusionAI-ing microenvironment of nodal metastasis enhances robustness of ECE prediction in patients with oropharyngeal cancer.