Interpretable dynamic quantitative vascular morphometry features using SHAP for anti-angiogenic therapy response prediction

Clinical Snapshot

Key Findings

• P Value: Not reported in abstract

• Effect Size: AUC-ROC = 0.842 (internal validation, fivefold cross-validation); AUC-ROC = 0.806 (external validation) for the delta-merge model

• Primary Outcome: Prediction of anti-angiogenic therapy response in lung cancer patients using delta quantitative vascular morphometry features from contrast-enhanced CT

• Nnt Or Sensitivity: Sensitivity and specificity not reported; only AUC-ROC provided. Likelihood ratios cannot be calculated from available data. SHAP analysis identified arterial vessel involvement and venous recovery as the dominant discriminating features.

• Confidence Interval: Not reported in abstract or available data

Clinical Application

The automated pipeline using routine contrast-enhanced CT is theoretically feasible in centres with appropriate computational infrastructure. However, the segmentation software, CT acquisition protocols, and computational requirements are not described in sufficient detail for immediate clinical translation. Validation in prospective, multicentre cohorts with diverse CT scanners and protocols is required before clinical deployment. Anti-angiogenic agents relevant to Australian practice include bevacizumab (PBS-listed for non-squamous NSCLC in combination with chemotherapy), ramucirumab, and nintedanib. The RACGP and relevant oncology guidelines (COSA, Cancer Council Australia) emphasise RECIST-based response assessment as the current standard. This tool, if validated, could complement RECIST by providing earlier or more granular vascular response data. However, TGA approval and prospective Australian validation would be prerequisites for clinical adoption. CT protocols in Australian radiology departments may differ from those used in this Chinese single-centre study, potentially affecting segmentation performance and feature reproducibility. PBS implications are indirect — improved early response prediction could inform continuation or cessation of costly anti-angiogenic agents. Adult patients with lung cancer (histological subtypes unspecified) being considered for or currently receiving anti-angiogenic therapy, with access to contrast-enhanced CT imaging at baseline and follow-up

Abstract

Anti-angiogenic therapy benefits vary, with response rates of 40 to 70%, highlighting the need for early biomarkers to identify responders. We developed an automated machine learning framework that uses delta quantitative vascular morphometry features from standard contrast-enhanced CT to evaluate treatment response. This workflow combines automated tumor and vessel segmentation with feature extraction from routine scans for clinical use. Shapley additive explanations (SHAP)-based attributions identify key vascular and clinical features, providing meaningful, imaging-visible evidence aligned with therapy targets beyond traditional radiomics. Using baseline and follow-up CTs from 163 patients with lung cancer, we built three models using fivefold cross-validation, with the delta-merge model achieving high accuracy (area under the receiver operating characteristic curve = 0.842 internally, 0.806 externally). SHAP analysis uncovered an "arterial-dominant, venous-adaptive" pattern, where arterial involvement and venous recovery distinguish responders. This automated workflow and visualization support early, imaging-based response assessment and personalized treatment.

References

1. 1.Hu, K., Cai, Q., Xu, J., Ai, S., Ou, W., & Liu, Y. (2026). Interpretable dynamic quantitative vascular morphometry features using SHAP for anti-angiogenic therapy response prediction. Science Advances. https://doi.org/10.1101/2024.11.19.24317538