Cell & Developmental Biology
Open Access

Abstract

RETRACTED: The Use of Molecular Diagnostics and Artificial Intelligence in the Classification and Prognosis of Gliomas: A Systematic Review and Meta-Analysis

Costas Fantis

The integration of molecular diagnostics and Artificial Intelligence (AI) algorithms has emerged as a potential approach in the classification and prognosis of gliomas, providing valuable insights to clinicians and researchers for improved patient care. This systematic review aimed to critically assess the current state of research on the utilization of molecular diagnostics and AI in glioma management. A comprehensive search strategy was conducted in the Cochrane Library, Embase, and PubMed databases, yielding 8,701 initial papers. After duplicate removal and screening by independent reviewers, 1,653 papers were included for further evaluation. Ultimately, 70 papers met all inclusion criteria and were subjected to data extraction, synthesis, and quality assessment. The systematic review presented key findings through three comprehensive tables and one informative pie chart. The majority of studies (85.7%, 65 papers) explored the impact of machine learning algorithms on molecular diagnostics, improving the accuracy and effectiveness of glioma classification and prognosis. A smaller proportion of studies (21.4%, 18 papers) provided information on integrating molecular diagnostics and AI for personalized treatment plans, including the identification of optimal drug combinations and targeted therapies. Machine learning algorithms demonstrated their potential in quantitative image feature analysis, deep learning based radiomics, and enhancing survival prediction. Various molecular biomarkers were identified in the systematic review, including IDH mutations, 1p/19q co-deletions, CDKN2A/B homozygous deletion, Methylguanine-DNA Methyltransferase (MGMT) promoter methylation, Estimated glomerular filtration rate (EGFR) alterations, TERT promoter mutations, and H3F3A and IDH mutations in pediatric gliomas, among others. Machine learning algorithms effectively analyzed genomic data and predicted the presence of these biomarkers, facilitating accurate glioma classification and personalized treatment planning. Despite potential results, challenges and limitations need to be addressed for successful implementation in clinical practice. These include the need for robust and high-quality data, rigorous validation, improved interpretability and transparency of AI models, cost-effectiveness, specialized expertise, and evolving regulatory frameworks. Recommendations were proposed, such as standardizing protocols, enhancing image quality, promoting explainable AI, multi-institutional collaboration, and developing user-friendly software. Future research should prioritize standardized protocols and guidelines, validation using external datasets, multiomics integration, and addressing ethical and legal considerations. By following these recommendations, researchers can bridge the gap between research and clinical application, leading to more effective and personalized approaches to diagnose, classify, and treat gliomas, ultimately improving patient outcomes. The findings from this systematic review highlight the transformative potential of molecular diagnostics and AI in glioma management and provide valuable insights into routine clinical practice.

Published Date: 2024-06-03; Received Date: 2024-05-02