Interface for interactive segmentation of magnetic resonance image

Abstract

This study was conducted to develop a user friendly computer interface that acts as a possible bridge between the gap of medical image viewing and the segmentation process. It aims to narrow down the time taken for segmentation process in knee magnetic resonance images (MRI) by providing a general shape for the region of interest (ROI). The ROI of the femoral articular cartilage was investigated more intensely as it is used by clinicians to diagnose the progress of osteoarthritis (OA). Other ROI being investigated includes the femur and the tibial articular cartilage. The idea purposed in this project was to use training data sets for any ROI that were prepared before hand to interpolate new data on medical images that has just been loaded onto the interface. The interpolated data and the medical images will be mapped together, and the user has the ability to correct any misalignment by using the functions of the spline widget provided within the interface. A total of 8 image volume sets were used and 2 types of runs between data sets were conducted. The first type involves image sets coming from the same patient taken at different timings, while the second type involves image sets coming from different patients. For the second type, it was further categorized into runs with patients having the same clinical conditions, and runs with patients having differing clinical conditions.