인문학
사회과학
자연과학
공학
의약학
농수해양학
예술체육학
복합학
DBpia AI
AI 에이전트
AI 아이디어
AI 리더
AI 뷰어
Citeasy
크롬 서지관리 다운로드
개인구독
소속 기관이 없으신 경우, 개인 정기구독을 하시면 저렴하게
논문을 무제한 열람 이용할 수 있어요.
지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
초록·키워드
Abstract Background Diffuse gliomas like glioblastoma, astrocytoma, and oligodendroglioma are defined by complex heterogeneity in imaging and molecular patterns. Traditional diagnosis relies severely on invasive biopsies and human interpretation of MRI scans, which are subjective and severely limited in capturing the complete volumetric and biological heterogeneity of the tumor. Artificial Intelligence (AI) can potentially enhance glioma tumor subtype classification through the analysis and interpretation of medical imaging information. Machine learning, a branch of AI, is key to discovering patterns and characteristics in the data to enable predictive models for prognosis and diagnosis. AI can process MRI scans to derive tumor shape, size, and texture information, and process large datasets to define risk factors and features. This method may improve the speed and accuracy of glioma subtype diagnoses and facilitate individualized treatment planning. Aim In the current study, we have evaluated the performance of multiple machine learning classifiers in distinguishing between glioblastoma, astrocytoma, and oligodendroglioma based on radiomics and clinical features. The goal of the study was to identify effective feature selection techniques that enhance the accuracy and reliability of classification models and to develop a methodology that can be adapted for use with imaging data collected at our institute, supporting future diagnostic workflows. Methods A total 729 radiomic features were expended from TCIA for the experimental analysis. The radiomic signature of significant features has been created for every sample using XGBoost decision tree, XGBoost random forest, CatBoost and Light GBM tree based feature selection algorithm. The experimentation has been done by training 13 different models. The hyper parameters of each of the model has been tuned and different performance parameters like accuracy, precision, recall, F1 score and AUC-ROC curve have been compared. Results Feature selection using XGBoost decision tree, XGBoost random forest, Catboost and Light GBM tree- based classifier is used for selecting the top 21, 37, 71 and 82 features based on feature importance score. The experimental results shows that the feature subset by XGBoost decision tree method gave the best performance as compared to others. A total of 13 classification models were trained and tested giving a best tenfold validation accuracy, test accuracy, macro F1-score of 95.2%, 77%, 0.716 and AUC of 0.92 for Catboost classifier.
인공지능 문자 인식 모델을 통해 추출된 텍스트로, 일부 오타나 오류가 포함될 수 있으나 지속적으로 개선 중입니다.
오류를 발견하셨다면 해당 부분을 드래그한 후 ' 를 통해 신고해주세요.
오류를 발견하셨다면 해당 부분을 드래그한 후 ' 를 통해 신고해주세요.
최근 본 자료 전체보기
