HUMAN BODY PARTS AND BONE RECOGNITION COMBINING 3D VISUALIZATION FOR DICOM IMAGES: CLASSIFICATION EXPLANATION AND 3D VISUALIZATION

Abstract

In this thesis, we have proposed a framework applying Machine Learning techniques to help doctors in medical diagnosis with the title: “Human Body Parts And Bone Recognition Combining 3D Visualization For Dicom Images: Classification Explanation and 3D visualization”. We focus on creating an application that can assist doctors in diagnosing with DICOM images as well as avoid mistakes when classifying them. In addition, we want to accurately mark the bone positions in parts of the human body im images. This helps doctors avoid confusion between bones and other areas and improve treatment efficiency. We also develop a website application that allows doctors to view medical images in 2D and visualize them into 3D models. To do so, we analyze the problems and difficulties of this study. Next, we have to learn the related knowledge, programming languages and libraries. After that we build and train the model. When the training is complete, we use the model to predict the human body parts and recognize the bones in them as well as test and evaluate the results. Finally, the web application is developed for viewing the result images. In conclusion, the classification results and the detection of bones are quite precise, the web application works smoothly and has stable performance. We believe that our proposal will surely help in the medical treatment and its application could be extended to be used in other fields.