Machine learning-based prognostic model for patients with anaplastic thyroid carcinoma

• DOI: 10.1007/s12672-024-01703-9
• Publication Date: 2025-01-21T13:23:27Z
• Authors: Yihan Sun; Da Lin; Xiangyang Deng; Yinlong Zhang
• Keywords: Anaplastic carcinoma

Despite the identification of various prognostic factors for anaplastic thyroid carcinoma (ATC) patients over years, a precise tool these is still lacking. This study aimed to develop and validate model predicting survival outcomes ATC using random forests (RSF), machine learning algorithm. A total 1222 were extracted from Surveillance, Epidemiology, End Results (SEER) database randomly divided into training set 855 validation 367 patients. We developed an RSF traditional Cox cohort further compared their performance based on calibration discrimination. integrated brier score (iBS) was used estimate ability. The Brier score, C-index value, receiver operating characteristic (ROC) curve with area under (AUC) Decision Curves Analysis (DCA) evaluated. Furthermore, we assessed feature importance within validated its group. An successfully in set. 0.055, which lower than model's 0.063, indicating better since signifies superior accuracy. exceeded AUC. Additionally, DCA indicated that provided substantial clinical benefit. And ranked time-dependent features according permutation observed surgery, radiotherapy, chemotherapy most influential predictors initially. Moreover, predictions, stratified 2 groups displaying significant difference survival. first revealed offers more overall predictions stratification regression