Purpose: Human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (OPSCC) can be treated with definitive radiotherapy (RT), combined chemoradiotherapy (CRT), and/or transoral robotic surgery (TORS). TORS is a surgical approach that provides a comparable overall survival (OS) rate of approximately 90% at 2 years with the potential for less toxicity, lower cost, and shorter treatment regimen. However, approximately one-third of TORS patients require adjuvant CRT (with the highest combined treatment toxicity profile) due to extranodal extension (ENE) or positive surgical margins on the final pathology. In retrospect, the decision for these patients to undergo TORS is unfavorable. This study describes a treatment management selection procedure which utilizes artificial intelligence (AI), applied to clinical and imaging data, such that it identifies the patients contraindicated for TORS.
Methods: Electronic medical records (EMR) for 55 subjects were retrospectively reviewed. Patient data included demographics, history, vital signs, and labs. Radiomics from diagnostic CTs were also utilized. After statistical pre-processing and remapping to equivalent numerical representations, three statistically significant textures were added to the analyses. Extracted data were used as an input into a deep neural network (DNN) that computed a probabilistic value for each patient, indicating whether TORS is a suitable disease management option. The DNN (based on Google’s TensorFlow) incorporated a supervised learning technique used to mimic the physician’s decision-making. A 5-fold cross-validation was utilized for evaluating the model’s performance.
Results: The DNN was successfully implemented with variable parameters, allowing flexibility and easy reconfigurability. An average AUC of 0.89 was achieved using significant EMR and imaging data.
Conclusion: Treatment selection of HPV-associated OPSCC patients may be facilitated by AI techniques, applied to commonly tallied clinical variables, and combined with diagnostic imaging radiomics. The proposed methodology will help in discriminating patients who would ultimately require tri-modality therapy with the highest toxicity profile.
Artificial Intelligence, Radiomics, EMR