Forecasting mutations in the telomerase reverse transcriptase promoter associated with glioma: A comprehensive systematic review and diagnostic meta-analysis utilizing machine learning algorithms

Background: Glioma represents one of the most prevalent forms of primary brain tumors. The mutation in the telomerase reverse transcriptase promoter (pTERT) is correlated with an improved prognosis. This research aims to explore the TERT mutation in glioma patients by employing machine learning (ML) algorithms on radiographic images. Method: The study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A comprehensive search of electronic databases, including PubMed, Embase, Scopus, and Web of Science, was performed from their inception until August ۱, ۲۰۲۳. Statistical analyses were executed using the MIDAS package within STATA version ۱۷. Results: A total of ۲۲ studies encompassing ۵,۳۷۱ patients were selected for data extraction, with data synthesis derived from ۱۱ reports. The analysis indicated a pooled sensitivity of ۰.۸۶ (۹۵% CI: ۰.۷۸–۰.۹۲) and a specificity of ۰.۸۰ (۹۵% CI: ۰.۷۲–۰.۸۶). The positive and negative likelihood ratios were calculated at ۴.۲۳ (۹۵% CI: ۲.۹۹–۵.۹۹) and ۰.۱۸ (۹۵% CI: ۰.۱۱–۰.۲۹), respectively. The pooled diagnostic score was determined to be ۳.۱۸ (۹۵% CI: ۲.۴۵–۳.۹۱), with a diagnostic odds ratio of ۲۴.۰۸ (۹۵% CI: ۱۱.۶۳–۴۹.۸۷). The Summary Receiver Operating Characteristic (SROC) curve exhibited an area under the curve (AUC) of ۰.۸۹ (۹۵% CI: ۰.۸۶–۰.۹۱). Conclusion: The findings indicate that machine learning (ML) has the potential to serve as an effective instrument for predicting TERT mutation status in individuals diagnosed with gliomas. The analysis revealed that ML models demonstrated a sensitivity of ۰.۸۶ and a moderate specificity of ۰.۸۰, which may aid healthcare practitioners in assessing disease prognosis and formulating suitable treatment strategies. Despite the encouraging outcomes reported in recent years concerning the diagnostic efficacy of ML models for predicting gene mutation status in glioma patients, the integration of these models into routine clinical practice remains constrained. There is a need for further refinement and enhancement of ML models to achieve improved performance metrics, thereby increasing their reliability and utility in clinical settings.

کلیدواژه ها:

artificial intelligence ، Machine learning ، deep learning ، telomerase reverse transcriptase ، glioblastoma