Computer Vision Approach for Liver Tumor Classification Using CT Dataset

The liver tumor is one of the most foremost critical causes death in world. Nowadays, Medical Imaging (MI) prominent Computer Vision fields (CV), which helps physicians and radiologists to detect diagnose tumors at an early stage. Radiologists use manual or semi-automated systems read hundreds images, such as Computed Tomography (CT) for diagnosis. Therefore, there a need fully-automated method using popular widely used imaging modality, CT images. proposed work focuses on Machine Learning (ML) methods: Random Forest (RF), J48, Logistic Model Tree (LMT), (RT) with multiple automated Region Interest (ROI) multiclass classification. dataset comprises four classes: hemangioma, cyst, hepatocellular carcinoma, metastasis. Converted images into gray-scale, contrast was improved by applying histogram equalization. noise reduced Gabor filter, image quality sharpening algorithm. Furthermore, 55 features were acquired each ROI different pixel dimensions texture, binary, rotational, scalability, translational (RST) techniques. correlation-based feature selection (CFS) technique deployed obtain 20 optimized from these results showed that RF RT performed better than J48 LMT, accuracy 97.48% 97.08%, respectively. novel framework will help tumors.