Artificial Intelligence Assisted Diagnosis of Cancer from Whole Body Bone Scan Images

Abstract

Decision making perception knowledge and planning are the fundamental prerequisite of tumor detection through artificial intelligence. To achieve these objectives, a variety of techniques are used, which may include Dicom conversion, mask RCNN and binary classifier. Diagnosis of cancer from whole bone scan image is divided in some algorithms. In the initial stage image is in the form of dicom which have information as well as image of the body. Image is converted in jpg format. Then for diagnosis of cancer from body scan image is converted in small body regions which have most targeted areas like skull chest and pelvis. This is done by using mask RCNN. Mask RCNN extract body regions from whole body bone scan image. All regions have its own model file when new image put in the algorithm then model weight files are used for extraction and save in the folder respectively. Then these images are used for binary classification. The classifier is self-make classifier for the classification of body regions. Every image has its own image size. Every image has different dimension due to body parts which are not in same size. The binary classifier model trained by using these images. For every body region model trained for classification. In this classification two classes are present normal and infected(metastatic). This will resolve the issue of ecologist mal diagnosis about metastasis. This is very useful for those patients which are at 4th stage of cancer. But this algorithm helps them to live more life by taking some precautions. And in Pakistan where number of patients are present but due to less availability of ecologists give them a way by which they can take some precautionary measures.