<![CDATA[This study presents an in-depth analysis of the application of a Decision Tree classifier to detect bone fractures from X-ray images, leveraging the FracAtlas dataset containing 4,083 labelled images. The classifier underwent a rigorous evaluation using 5-fold cross-validation, focusing on metrics such as accuracy, precision, recall, and F1-score to ascertain its performance. Results varied across folds, with an accuracy range of 69.89% to 74.05%, precision between 72.27% and 73.75%, recall from 70.50% to 73.81%, and F1-scores of 71.52% to 73.31%. A graphical depiction of these metrics provided a visual comparison of performance consistency, while the confusion matrix offered a detailed account of the model’s predictive success and shortcomings. The research confirms the hypothesis that integrating Canny edge detection for segmentation and Hu Moments for feature extraction with a Decision Tree approach can facilitate fracture identification, positing the model as a supportive tool for radiologists. The study's findings contribute to the field of medical image analysis, suggesting that machine learning can be a valuable asset in clinical diagnostics. Recommendations for future research include the exploration of more complex algorithms, expansion of the dataset, and refinement of pre-processing techniques, to enhance the model's diagnostic precision further.]]>
Copyrights © 2024
