Optical imaging of treatment-naive human NSCLC reveals changes associated with metastatic recurrence

Abstract

Lung cancer remains the leading cause of cancer deaths, comprising nearly 25% of all cancer deaths [1]. NSCLC accounts for approximately 85% of all cases and encompasses major subtypes such as adenocarcinoma and squamous cell carcinoma. Despite advances in surgical and therapeutic options, NSCLC remains associated with poor prognosis due to a high rate of recurrence, even in early stages. Around 30-55% of patients who undergo complete resection will experience metastatic recurrence, significantly lowering survival outcomes [2]. There is a critical need to develop prognostic markers capable of predicting risk of recurrence at earlier timepoints in order to improve NSCLC management, as it could help clinicians tailor treatment plans, optimize follow-up schedules, and identify high-risk patients who might benefit from adjuvant therapies. Two photon microscopy (TPM) techniques provide non-invasive high-resolution information on cell metabolism within tissue by utilizing an optical redox ratio (ORR) of FAD/[NADH+FAD] autofluorescence. The goal of this study is to use the ORR and NADH fluorescence lifetime decay to identify measurable differences in optical endpoints of human NSCLC that are indicative of their long-term outcome. Twenty-five treatment-naive NSCLC specimens were classified into metastatic and non-metastatic groups according to subject-detail reports. The ORR and mean NADH lifetime were determined for each sample, revealing a significant increase in the ORR for the metastatic group. Additionally, tumors presenting with high optical redox ratios were found to be correlated with low KEAP1, a prognostic indicator of poor clinical outcome in NSCLC. To evaluate the prognostic potential of optical metabolic endpoints, we trained three classifiers: logistic regression, SVM, and KNN on three different feature sets: optical endpoints, clinicopathological features, and combination of optical and clinical features. We found that SVM trained on optical endpoints alone (AUC = 0.74) outperformed the model built with only clinical features (AUC = 0.62), when classifying tumors based on their metastatic recurrence status. Together, these findings highlight the potential of optical metabolic imaging to provide markers of recurrence in NSCLC.