Fig. 6: Construction and evaluation of the multi-feature predictive model. | npj Precision Oncology

Fig. 6: Construction and evaluation of the multi-feature predictive model.

From: Serial monitoring of cerebrospinal fluid in patients with leptomeningeal metastases in lung cancer via the Ommaya reservoir as a predictive indicator of therapeutic efficacy and clinical prognosis

Fig. 6

a Workflow of constructing the multi-feature model. b Comparative analysis of model accuracy using different feature combinations trained with the RF algorithm. c ROC curve of the RF model trained with all feature combinations in the validation cohort. d Confusion matrix of the RF model trained with the optimal feature combination in the validation cohort. Performance comparison of the optimal feature combination using the FNN algorithm: accuracy and F1 score (e), ROC curve (f), confusion matrix (g), and PR curve (h) for the feature combinations model in the validation cohort. Abbreviations: DFCs different feature combinations, FNN feed-forward Neural Network, NRE non-responses (SD/PD), PR curve, precision–recall curve, RE responses, RF random forest.

Back to article page