S657
Clinical – Head & neck
ESTRO 2026
planning Resource Constraints. Preprint. medRxiv. 2024;2024.04.01.24305163. Published 2024 Nov 4. doi:10.1101/2024.04.01.24305163 Keywords: AI, Large Language Model, Adaptive Radiotherapy Digital Poster 4375 Histopathological Predictors of Survival in Adjuvant Oral Cavity Squamous Cell Carcinoma: A Retrospective Analysis Pratiksha Tyagi 1 , Sarbani Ghosh Laskar 1 , Subramanya Adiga 1 , Samarpita Mohanty 1 , Anuj Kumar 1 , Shwetabh Sinha 1 , Ashwini Budrukkar 1 , Monali Swain 1 , Munita Bal 2 , Asawari Patil 3 , Neha Mittal 2 , Swapnil Rane 2 , Gouri Pantvaidya 4 , Shiva Kumar Thiagarajan 4 , Chandra Shekhar Dravid 4 , Anuja Deshmukh 4 , Deepa Nair 4 , Vidisha Tuljapurkar 4 , Rukmini Prabhu 4 , Sudhir Nair 4 , Poonam Joshi 4 , Rathan Shetty 4 , Arjun Singh 4 , Pankaj Chaturvedi 4 1 Radiation Oncology, Tata Memorial Hospital, Mumbai, India. 2 Department of Pathology, Tata Memorial Hospital, Mumbai, India. 3 department of Pathology, Tata Memorial Hospital, Mumbai, India. 4 Department of Surgical Oncology, Tata Memorial Hospital, Mumbai, India Purpose/Objective: Oral cavity squamous cell carcinoma (OCSCC) id prognostically affected by various clinicopathological characteristics. Even with advances in the treatment modalities, the outcomes are poor for patients with high risk histopathological features like extranodal extension (ENE), depth of invasion (DoI), pattern of invasion (WPOI/PPOI), perineural invasion (PNI), lymphovascular invasion (LVI) and cut margin status. The study aims on correlating these features with outcomes and patterns of failure. Material/Methods: A retrospective analysis of 1186 patients of OCSCC was conducted between January 2020 and April 2025, at a tertiary care center in Western India. Median follow up was 20 months. Statistics were analysed using Kaplan- Meier, log-rank, Chi-square, Fisher’s exact, and Cox proportional hazards testson SPSS v 26. Results: Most common tumor sites were buccal mucosa (37.7%) and lateral tongue (33.9%).Adverse features were WPOI Type 4 (67.5%), PPOI Type 3 (68.9%), LVI (6.3%) and PNI (39.3%).Margins were involved (<1mm) in 2% and close (1–5mm) in 16.4%; involved margin significantly associated with LVI (p=0.025).Most patients were pT4a (38%), pN0 (36.4%) and DOI >1mm (55.1%).pT status was significantly associated with WPOI type 5 (p=0.048) and PNI (p<0.01).ENE was present in 36.1%, with 9.6% major and 22.7% minor
supporting literature citations. The results demonstrated that during the RT duration in a represented patient, the trigger score progressively increased. The LLM's triggering decision incorporates not only anatomical and dosimetric information but also patient-specific factors such as treatment-related toxicities and weight loss during radiotherapy. Following clinician review of the LLM’s recommendation, multiple ART interventions were implemented for this test-case patient.
Conclusion: We have successfully developed an automated, evidence-based ART-triggered decision support system in NPC and confirmed its operational feasibility. The system provides real-time assessments for triggering ART during a patient's radiotherapy, potentially saving time and minimize inter-physician variability in decision-making. Future work will focus on conducting large-scale prospective clinical validations to assess the system's actual impact on clinical outcomes. References: 1.Lv J, Xu LX, Li ZX, et al. Longitudinal on-treatment circulating tumor DNA as a biomarker for real-time dynamic risk monitoring in cancer patients: The EP- SEASON study. Cancer Cell. 2024;42(8):1401-1414.e4. doi:10.1016/j.ccell.2024.07.0012. Wu PM, Chua DT, Sham JS, et al. Tumor control probability of nasopharyngeal carcinoma: a comparison of different mathematical models. Int J Radiat Oncol Biol Phys. 1997;37(4):913-920. doi:10.1016/s0360-3016(96)00588- 33. Nosrat F, Dede C, McCullum LB, et al. Optimal Timing of Organs-at-Risk-Sparing Adaptive Radiation Therapy for Head- and-Neck Cancer under Re-
Made with FlippingBook - Share PDF online