Hierarchical clustering-based segmentation (HCS) aided diagstic image interpretation monitoring.

Machines are good at operations which require precision and computing objective measures. In contrast, humans are good at generalisation and making decisions based on their past experience and heuristics. Hence, to solve any problem with a solution involving human-machine interaction, it is imperative that the tasks are shared appropriately. However, the boundary which divides these two different set of tasks is not well defined in domains such as medical image interpretation. Therefore, one needs a versatile tool which is flexible enough to accommodate the varied requirements of the user. The aim of this study is to design and implement such a software tool to aid the radiologists in the interpretation of diagnostic images.Tissue abnormality in a medical image is usually related to a dissimilar part of an otherwise homogeneous image. The dissimilarity may be subtle or strong depending on the medical modality and the type of abnormal tissue. Hierarchical Clustering-based Segmentation (HCS) process is a dissimilarity highlighting process that yields a hierarchy of segmentation results. In this study, the HCS process was investigated for offering the user a versatile and flexible environment to perceive the varied dissimilarities that might be present in diagnostic images. Consequently, the user derives the maximum benefit from the computational capability (perception) of the machine and at the same time incorporate their own decision process (interpretation) at the appropriate places.As a result of the above investigation, this study demonstrates how HCS process can be used to aid radiologists in their interpretive tasks. Specifically this study has designed the following HCS process aided diagnostic image interpretation applications: interpretation of computed tomography (CT) images of the lungs to quantitatively measure the dimensions of the airways and the accompanying blood vessels; Interpretation of X-ray mammograms to quantitatively differentiate benign from malignant abnormalities. One of the major contribution of this study is to demonstrate how the above HCS process aided interpretation of diagnostic images can be used to monitor disease conditions. This thesis details the development and evaluation of the novel computer aided monitoring (CAM) system. The designed CAM system is used to objectively measure the properties of suspected abnormal areas in the CT images of the lungs and in X-ray mammogram. Thus, the CAM system can be used to assist the clinician to objectively monitor the abnormality. For instance, its response to treatment and consequently its prognosis. The implemented CAM system to monitor abnormalities in X-ray mammograms is briefly described below. Using the approximate location and size of the abnormality, obtained from the user, the HCS process automatically identifies the more appropriate boundaries of the different regions within a region of interest (ROI), centred at the approximate location. From the set of, HCS process segmented, regions the user identifies the regions which most likely represent the abnormality and the healthy areas. Subsequently, the CAM system compares the characteristics of the user identified abnormal region with that of the healthy region; to differentiate malignant from benign abnormality. In processing sixteen mammograms, the designed CAM system demonstrated the possibility of successfully differentiating malignant from benign abnormalities.