Bone Tumor Segmentation from CT Images

Computer Aided Diagnosis (CAD) is one of the important topics of radiology and medical image processing. CAD systems brings second opinion to physicians by using algorithms and related software for diagnosis and treatments. In this study, an automatic and semi-automatic bone tumor extraction methods have been proposed. Bone tumors have been extracted from CT images. Proposed study consists of five steps which are; (i) pre-processing, (ii) segmentation, (iii) morphologic filtering, (iv) 3D modelling, (v) analysis and interpretation. MATLAB platform and 3D-Doctor software have been used for 2D and 3D processing respectively. This study has been realized by using retrospective nine CT data (male-female). Obtained results have been validated with clinical findings of radiology and orthopedic experts. Accuracy assessment of the study has been done by ROC curve. As a result of the study, it has been observed that besides increasing accuracy of diagnosis, CAD systems are capable to prevent misdiagnosis due to heavy working conditions and lack of concentration of physicians. Distinguish between bone tumors and tissues which resembling bone tumors is quite difficult. Therefore, it is thought that, proposed methods in this study can contribute other studies for diagnosis and treatment purposes, especially for cancerous lesions.