An Expedition to Intelligent Diagnosis of Bone Cancer and It's Direction from Capsule Network
Md Abid Hassan, Md. Hasibul Hasan, Oyadud Chowdhuri, Md Ariful Haque, Md. Siyamul Islam · 2022
Although Early diagnosis is the lifesaving strategy and superfluous appurtenant to bone cancer. Pathologists, however, find it incommodious to detect bone cancer early due to cellular heterogeneity in Computerized Tomography (CT) and Magnetic Resonance Imaging (MRI) images. In this regard, many image processing techniques can assist doctors in classifying cancer and non-cancer samples intelligently. Among various image processing and data mining techniques, Deep Learning outperforms other techniques in image classification applications. However, popular Deep Learning architectures (i.e. AlexNet, VGG-16, Inception-Net, ResNet etc.) still have some serious drawbacks where higher level layers combine lower level features and then construct result regardless of order and spatial relations among the image features. Moreover, it exhibits less accuracy with small dataset in bone cancer classification. This research aims to classify bone cancer from MRI and CT images with high accuracy and less training data as there is scarcity of real bone cancer data. Therefore, we propose an approach combining with Capsule Network which satisfies the research aim. The qualitative and numerical analysis exhibits that, proposed framework provides upper horizontal accuracy than other Deep Learning techniques with the percentage of 95.26 while detecting the bone cancer.