Show simple item record

AdvisorAbdallah, Atiyeh
AuthorMoosa, Esra Mohamed
Available date2024-02-05T07:29:04Z
Publication Date2024-01
URIhttp://hdl.handle.net/10576/51523
AbstractBackground: Accurately diagnosing CNS tumors is essential but difficult, with a notable risk of errors. To enhance diagnostic precision, the WHO now combines molecular data with traditional methods. While AI and ML technologies show promise for improving medical diagnosis, integrating them into clinical practice is complex. This study aims to investigates and assesses the effectiveness of the DKFZ DNA methylation classifier for pediatric CNS tumors and explores the necessary requirements for its adoption in clinical settings. Methods: This retrospective study included all pediatric patients under 16 who were diagnosed with CNS cancer at Sidra Hospital in Qatar from January 2018 to April 2023, and it also included CNS cases from Turkey between 1996 and 2020. The samples undergo DNA extraction, DNA methylation profiling, and classification of each tumor using the MNP brain tumor classifier. The original histopathological assessments were then cross-checked with the classifications derived from DNA methylation. In cases of mismatched results, an expert neuropathologist re-examined the findings. Findings: Variations in array run quality were evident, particularly in the run dated 2023-03-13, which showed lower CpG counts and consequent classification scores. This highlighted the necessity of strict adherence to protocols to ensure consistency and accuracy of results. Importantly, methylation profiling significantly refined initial diagnoses in 73.3% of cases from Sidra, demonstrating the technique's value in informed treatment decision-making. Specifically, medulloblastoma diagnoses benefited from the classifier, showing that even with lower-quality DNA, the classifier could provide high match rates for the diagnosis. Conclusion: In summary, this study confirms that DNA methylation profiling can effectively support the diagnosis of CNS tumors in young patients and shows reliable performance in two cohort groups. This method is reliable, precise, and valuable for categorizing pediatric CNS tumors, especially for detailed subtyping of medulloblastomas. To effectively adopt the DNA methylation classifier in clinical settings, it's essential to maintain high-quality standards and rigorous procedures to achieve accurate classifications and high matching scores.
Languageen
SubjectDNA
Machine Learning
Artificial Intelligence
TitleQuality Control and Regulatory Framework for Implementing Machine Learning in Clinical Diagnosis: The DNA Methylation Classifier in Pediatric Brain Tumors
TypeMaster Thesis
DepartmentBiomedical Sciences
dc.accessType Full Text


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record