Integration of Node-RADS and Habitat Radiomics for Predicting Occult Nodal Metastasis in Bladder Cancer

This study aimed to evaluate the role of tumor habitat analysis in predicting occult nodal metastasis (ONM) in bladder cancer patients.

This retrospective study enrolled 84 bladder cancer patients who underwent pelvic lymphadenectomy. The tumor area was segmented into sub-regions using a Fuzzy C-Means clustering algorithm based on Hounsfield unit (HU) values. Independent clinical factors related to ONM were screened through univariate logistic regression analysis. Radiomics features were extracted from both the whole tumor and each sub-region to construct predictive models. The predictive performance of different models was evaluated. Furthermore, a nomogram was constructed by integrating habitat radscore with clinical factors, and its predictive efficacy was evaluated using the area under the curve (AUC). Finally, the practical clinical value of the nomogram was systematically evaluated through 1000 bootstrap iterations combined with decision curve analysis.

The whole-tumor model showed moderate discrimination (AUC: 0.735 training, 0.727 validation). The best single-habitat model (region 3) achieved AUCs of 0.859 and 0.788, whereas the regions 1, 2, and 3 model reached 0.920 and 0.864. The nomogram yielded AUCs of 0.952 (training) and 0.742 (validation).

Habitat radiomics improves ONM prediction over whole-tumor radiomics, and an integrated nomogram shows promising clinical utility pending external validation.