# From Radical Resection to Precision Surgery: Integrating Diagnostic Biomarkers, Radiomics-Based Predictive Models, and Perioperative Systemic Therapy in Head and Neck Oncology

## TL;DR

This paper reviews how head and neck cancer surgery has evolved to integrate diagnostic biomarkers and predictive models for more precise treatment and better outcomes.

## Contribution

The paper highlights the integration of genomic profiling, radiomics, and predictive models to enable precision surgery and adaptive treatment strategies.

## Key findings

- Comprehensive genomic profiling identifies actionable alterations in over 90% of head and neck squamous cell carcinomas.
- Radiomics-based models using machine learning achieve over 85% accuracy in predicting treatment response.
- Integration of these diagnostics allows response-adaptive strategies, preserving function in HPV-associated cases.

## Abstract

Head and neck cancer surgery has evolved from radical organ-sacrificing procedures to function-preserving approaches integrated within multidisciplinary frameworks. This comprehensive literature review, concentrating on studies from the past five years while incorporating relevant publications from the last three decades and landmark historical papers, examines the evolving role of surgery emphasizing diagnostic methodologies including comprehensive genomic profiling, validated imaging biomarkers, and their clinical integration for treatment selection and response prediction. Modern surgical practice demonstrates a paradigm shift toward precision medicine through validated diagnostic technologies. Comprehensive genomic profiling identifies clinically actionable alterations in over 90% of head and neck squamous cell carcinomas, with tumor mutational burden serving as a validated predictive biomarker for immunotherapy response. Programmed death-ligand 1 (PD-L1) combined positive score functions as a validated diagnostic biomarker for immunotherapy efficacy, demonstrating significant clinical benefit in biomarker-selected populations. Radiomics-based predictive models utilizing machine learning algorithms achieve diagnostic accuracies exceeding 85% for treatment response prediction when validated across independent cohorts. Quantitative ultrasound spectroscopy combined with magnetic resonance imaging radiomics demonstrates high sensitivity and specificity for radiation response prediction. Habitat imaging techniques characterizing tumor microenvironmental heterogeneity predict pathologic complete response to neoadjuvant chemoimmunotherapy with area under the curve values approaching 0.90 in validation studies. Integration of these diagnostic methodologies enables response-adaptive treatment strategies, with neoadjuvant chemotherapy facilitating mandibular preservation and adjuvant therapy omission in over half of human papillomavirus (HPV)-associated cases following surgical downstaging. Clinical validation of these diagnostic platforms enables accurate treatment response prediction and informed surgical decision-making, though standardization across institutions and demonstration of survival benefits through prospective trials remain essential for broader implementation.