Structured Proxy Features for Multimodal NSCLC Survival Prediction from Pretreatment CT

Lung cancer results in roughly 1.8 million fatalities annually worldwide, with Non-small cell lung cancer (NSCLC) comprising the majority of cases. Despite advancements in treatment, survival stratification remains challenging due to intratumoral heterogeneity inadequately captured by conventional descriptors. Standard radiomic and deep learning techniques regard imaging features as independent quantities, overlooking structured interactions between tumor characteristics. We evaluate whether structured proxy features can enhance multimodal NSCLC survival prediction by augmenting pretreatment (Computed Tomography) CT representations, radiomics, and clinical variables with six simulation-derived features designed to capture interactions between heterogeneity and morphology. A radiomics-parameterized cellular automaton turns entropy and sphericity into low-dimensional proxy parameters, which then create growth rate and necrosis ratio as structured features. The imaging backbone is a Transformer-based Masked Autoencoder (TMAE), which was chosen after a systematic evaluation with alternative encoders within the same pipeline and provides spatially explainable attention maps that identify tumor regions most informative for prognosis. On the public Lung1 cohort (n = 390), the primary four-modality fusion attained a C-index of 0.641 (iAUC 0.731, log-rank p < 0.001). In a distinct coefficient-optimization study, the best configuration achieved a C-index of 0.662 (iAUC 0.748). Both exceed prior multimodal results on Lung1 (C-index 0.631; iAUC 0.592 [15]) utilizing comparable evaluation protocols. These findings indicate that proxy features may offer complementary prognostic insights beyond standard radiomic, deep, and clinical representations within this benchmark framework; however, their robustness across various splits and external cohorts has yet to be confirmed. The current findings support retrospective feasibility for prognostic ranking derived from routinely acquired pretreatment imaging.