Study on computer-aided diagnosis of hepatic MR imaging and mammography
- Electronics and Information Systems Engineering Division, Graduate School of Engineering, Gifu University, 1-1 Yanagido, Gifu 501-1193 (Japan)
It is well known that the liver is an organ easily attacked by diseases. The purpose of this study is to develop a computer-aided diagnosis (CAD) scheme for helping radiologists to differentiate hepatic diseases more efficiently. Our software named LIVERANN integrated the magnetic resonance (MR) imaging findings with different pulse sequences to classify the five categories of hepatic diseases by using the artificial neural network (ANN) method. The intensity and homogeneity within the region of interest (ROI) delineated by a radiologist were automatically calculated to obtain numerical data by the program for input signals to the ANN. Outputs were the five pathological categories of hepatic diseases (hepatic cyst, hepatocellular carcinoma, dysplasia in cirrhosis, cavernous hemangioma, and metastasis). The experiment demonstrated a testing accuracy of 93% from 80 patients. In order to differentiate the cirrhosis from normal liver, the volume ratio of left to whole (LTW) was proposed to quantify the degree of cirrhosis by three-dimensional (3D) volume analysis. The liver region was firstly extracted from computed tomography (CT) or MR slices based on edge detection algorithms, and then separated into left lobe and right lobe by the hepatic umbilical fissure. The volume ratio of these two parts showed that the LTW ratio in the liver was significantly improved in the differentiation performance, with (25.6%{+-}4.3%) in cirrhosis versus the normal liver (16.4%{+-}5.4%). In addition, the application of the ANN method for detecting clustered microcalcifications in masses on mammograms was described here as well. A new structural ANN, so-called a shift-invariant artificial neural network (SIANN), was integrated with our triple-ring filter (TRF) method in our CAD system. As the result, the sensitivity of detecting clusters was improved from 90% by our previous TRF method to 95% by using both SIANN and TRF.
- OSTI ID:
- 20634660
- Journal Information:
- Medical Physics, Vol. 32, Issue 4; Other Information: DOI: 10.1118/1.1884765; (c) 2005 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA); ISSN 0094-2405
- Country of Publication:
- United States
- Language:
- English
Similar Records
Investigation of psychophysical similarity measures for selection of similar images in the diagnosis of clustered microcalcifications on mammograms
The simulation of 3D microcalcification clusters in 2D digital mammography and breast tomosynthesis