1269P - Machine learning model for predicting lung cancer recurrence after surgical treatment: A retrospective study using NLST and European hospital data

Background

The rate of lung cancer recurrence following curative surgical resection is 30-55% and remains a significant challenge in patient management. Accurate prediction of recurrence risk is crucial for guiding treatment decisions, such as the use of (neo-)adjuvant chemo- or immunotherapy, the extent of lung resection, and follow-up strategies. We present a preoperative machine learning model that uses patient computed tomography (CT) images and demographic features to predict lung cancer recurrence.

Methods

We collected a dataset of 588 clinical stage I-IIIA lung cancer patients who underwent surgical treatment for lung cancer, of which 147 had lung cancer recurrence. This includes retrospectively collected CT images and associated demographic and pathological data from patients in both screening and clinical settings from the US National Lung Screening Trial (NLST) and the North Estonia Medical Centre (NEMC). The preoperative model was trained to predict the likelihood of recurrence on a diverse set of features, including radiomic features extracted from CT images and relevant clinical variables. An 8-fold cross validation strategy was used, where, in each fold, 6 (of the 8 equally sized) subsets were used for training, one for model tuning, and one for validation. As a baseline, we compare the preoperative model to ranked clinical staging. Performance was evaluated using the Area-Under-the-ROC-Curve (AUC), sensitivity, and specificity.

Results

Lung cancer recurrence prediction results are tabulated below. We find that our model performs significantly better (AUC=61.4, Sen=18.2) than preoperative staging alone (AUC=56.1, Sen=16.4, p-value=0.035). Table: 1269P

AUC, sensitivity and specificity performance of our machine learning model compared with ranked TNM staging. The specificity is set to 90.0 for a rule-in context

Conclusions

Based on this retrospective analysis, we find that our model outperforms clinical staging prediction of lung cancer recurrence in preoperative settings. With further development, this algorithm could prove a valuable tool to aid the management of lung cancer patients.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Optellum.

Funding

Optellum.

Disclosure

A. Gasimova, N. Waterfield Price, L. Freitag: Financial Interests, Personal, Full or part-time Employment: Optellum. All other authors have declared no conflicts of interest.