Deep-learning-based survival prediction of patients with lower limb melanoma

• DOI: 10.1007/s12672-023-00823-y
• Publication Date: 2023-11-30T04:28:28Z
• Authors: Jinrong Zhang; Hai Yu; Xinkai Zheng; Wai-kit Ming; Yau Sun Lak; Kong Ching Tom; Alice Lee; Hui Huang; Wenhui Chen; Jun Lyu; Liehua Deng
• Keywords: Concordance

Abstract Background For the purpose to examine lower limb melanoma (LLM) and its long-term survival rate, we used data from Surveillance, Epidemiology End Results (SEER) database. To estimate prognosis of LLM patients assess efficacy, a powerful deep learning neural network approach called DeepSurv. Methods We gathered on those who had an diagnosis between 2000 2019 SEER divided people into training testing cohorts at 7:3 ratio using random selection technique. likelihood that would survive, compared results DeepSurv model with Cox proportional-hazards (CoxPH) model. Calibration curves, time-dependent area under receiver operating characteristic curve (AUC), concordance index (C-index) were all how accurate predictions were. In this study, total 26,243 enrolled, 7873 serving as cohort 18,370 cohort. Significant correlations age, gender, AJCC stage, chemotherapy status, surgery regional lymph node removal outcomes found by CoxPH The model’s C-index was 0.766, which signifies good degree predicted accuracy. Additionally, created data, higher 0.852. addition calculating 3-, 5-, 8-year AUC values, predictive performance both models evaluated. equivalent values for 0.795, 0.767, 0.847, respectively. model, in comparison, better 0.872, 0.858, 0.847. comparison demonstrated greater prediction patients, shown calibration curve. Conclusion patient database, performed than predicting time patients.