Medical imaging can provide important information for clinical or surgical physicians in the diagnosis of the disease process or pre-operative assessment with computer-aided system (computer-aided diagnosis, CAD) application. It is able to interpret a medical image to data information.
The purpose of this study is to establish a semi-automatic computer-aided detection system for liver segmentation for liver volume measurement. The current liver volume segmentation is performed manually by physicians in our hospital (Chang Gung Memorial Hospital). It is time-consuming and the result is dependent on experiences which is hardly repeatable. Based on the developed computer-aided detection system, clinicians are able to access the ratio of liver volume more efficiently. We use CT images to perform live volume measurement. However, there are variations between ages, body sizes, and genders, which make different representations in CT images. Luckly the range of Hounsfield unit values in these liver images are similar, which are useful information applied in our system. The scheme can be divided into three parts: 1. Strengthen and remove the bone part. 2. Strengthen the determined edge of liver image. 3. Use region growing to group liver pixels so that the liver area can be obtained. Find the complete liver contour. Calculate the liver volume ratio. Finally, we use the TPF FPF accuracy rate, and error rate to be the evaluation. The results are 93. 2 ± 2%, 99. 02 ± 0. 24%, 98. 8 ± 0. 25%, 1. 12 ± 0. 25%, respectively.
