Research Information System
Voprosy Onkologii, vol.71, no.2, pp.344-352, 2025 (Scopus)
Aim. Survival prediction in brain cancer is critical for treatment planning and patient outcomes. This study aimed to develop a prognostic model for brain cancer survival using supervised machine learning approaches. The model integrated demographic, clinical, immunohistochemical, and genomic data. Materials and Methods. We retrospectively analyzed 149 patients with intracranial tumors who underwent surgery. Demographic and clinical data were systematically collected. Tumor and adjacent tissues underwent histopathological and immunohistochemical analysis for GST-P, GST-T, GSTM, CYP1A1, CYP1B1, MDR, and p53 expression. Genomic DNA from tumors was analyzed for GSTM1, GSTT1, and p53 genotypes. Models were developed using decision tree, Naïve Bayes, and SVM algorithms in Python. Models were compared based on accuracy, precision, sensitivity, and F-measure metrics. Results. The overall postoperative survival rate was 65 %. Significant differences in protein expression were observed between cancerous and normal tissues for GST-P, GST-T, GST-M, CYP1A1, CYP1B1, MDR, and p53. GST-M1 null genotype was associated with brain tumor development. The decision tree model achieved the highest accuracy (84 %) among models integrating demographic, clinical, immunohistochemical, and genetic data. Precision and sensitivity varied across models, with the decision tree showing acceptable performance. Conclusion. Decision tree models are effective for predicting brain cancer survival, especially with limited datasets, using demographic, clinical, immunohistochemical, and genotypic variables.