AI and Cancer Research ── Lung ── 2026-11-14
Lung cancer: treatment, trials and AI
- Abstract
This study aimed to develop and validate an interpretable machine learning (ML) model that integrates clinical data, normalized distance from the metabolic hotspot to the tumor centroid (NHOC) / normalized distance from the metabolic hotspot to the tumor perimeter (NHOP) metrics, and PET/CT habitat radiomics to predict recurrence in lung adenocarcinoma (LUAD) and perform postoperative risk stratification. In this retrospective study, we analyzed patients with resectable stage Ⅰ LUAD from two hospitals and a public dataset, comprising a training cohort ( n = 253) and an internal validation cohort ( n = 106), including a subset with radiogenomics data ( n = 26). NHOC and NHOP were extracted using LIFEx software. Nine machine learning algorithms were evaluated to identify the optimal radiomics features. The best-performing algorithm was used to construct predictive models incorporating clinical variable, NHOC/NHOP, and habitat-based radiomics features. Shapley Additive Explanations (SHAP) were employed to interpret model contributions, while paired RNA-sequencing explored potential biological pathways. Smoking history and NHOCpeak were identified as independent predictors constituting the clinical model. The combined model demonstrated moderate discrimination performance for recurrence prediction. achieved an AUC of 0.876 (95% CI: 0.824–0.929) in the training cohort (Sensitivity: 0.890, Specificity: 0.720, PPV: 0.565, NPV: 0.942) and an AUC of 0.794 (95% CI: 0.671–0.918) in the internal testing cohort (Sensitivity: 0.833, Specificity: 0.750, PPV: 0.568, NPV: 0.919). NHOCpeak was the highest-ranked individual predictor in the SHAP analysis, whereas RadScore summarized the aggregate contribution of the radiomic features. The hybrid model successfully stratified patients into distinct risk groups for recurrence-free survival (RFS). Notably, the high-risk group showed enrichment in pathways related to motor proteins and neuroactive ligand-receptor interactions. This study presents exploratory findings of an interpretable machine learning model with moderate discrimination for the postoperative prediction of recurrence in stage I LUAD. NHOC emerged as a promising, novel independent predictor, providing significant added prognostic value beyond classical PET parameters.
Journal IF-equivalent: 5.7 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Xue B, Ji W, Wang L, Zheng X, Wang L, Zhang S, et al. Interpretable PET/CT habitat radiomics integrates novel metabolic metrics for postoperative risk stratification in stage I lung adenocarcinoma. Cancer Imaging. 2026 Nov 10 [Epub ahead of print]. doi:10.1186/s40644-026-01158-x.Checked: Abstract only - Abstract
BACKGROUND: Histological grading of lung adenocarcinoma is limited by interobserver variability, particularly when distinguishing noninvasive lepidic growth from invasive patterns. This distinction depends on evaluation of the alveolar elastic fiber framework, which is often insufficiently visualized on routine hematoxylin and eosin (H&E) slides. METHODS: We developed a deep learning framework to improve grading objectivity by computationally visualizing elastic fiber-related architecture from standard H&E images. A dual-stream EfficientNet-B0 classifier was trained using spatially coregistered H&E and eosin-based elastin fluorescence (EBEF) images. To enable clinical deployment without additional staining or fluorescence microscopy, a Pix2Pix-HD generative adversarial network was trained to synthesize high-fidelity virtual EBEF images from H&E alone. The resulting automated pipeline classified histological patterns, quantified pattern proportions, and assigned tumor grades. Performance was evaluated on independent multicenter cohorts, with ablation analyses assessing key pipeline components. Prognostic relevance was explored using a curated TCGA lung adenocarcinoma cohort. RESULTS: The benchmark dual-stream model using real H&E and EBEF achieved 89.3% tile-level classification accuracy. Using virtual EBEF images, the automated grading pipeline achieved 94.2% accuracy on an external validation cohort, outperforming an H&E-only model (79.6%). Ablation studies confirmed that performance gains depended on informative multimodal input, accurate spatial registration, and attentive feature fusion. AI-assigned grades demonstrated a consistent trend toward association with overall survival in TCGA, with significant stratification observed when low-grade tumors were combined (log-rank p = 0.048). CONCLUSIONS: This framework enables objective grading of lung adenocarcinoma from a single routine H&E slide by enhancing visualization of elastic fiber-related architecture. The approach improves grading consistency and prognostic stratification without altering standard pathology workflows.
Journal IF-equivalent: 4.2 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Wang CL, Lei SP, Lu Y, Zou LF, Xiong JB, Li JW, et al. Deep learning-powered virtual eosin-based elastic fiber framework imaging improves objective grading and prognostic stratification of lung adenocarcinoma. Diagn Pathol. 2026 Aug 17;21(1):92. doi:10.1186/s13000-026-01837-0. PMID: 43015879.Checked: Abstract only