Pages 15-16
The use of the ultrasound-imaging platform in medical facilities has significantly increased in recent years due to its superior capability, namely high resolution, small influence to the human body and so on. Though the doctors diagnose a tissue characterization of a liver by observing an image obtained by ultrasound, a diagnostic result seems to make a difference with experience of doctors. Therefore a quantitative analysis method concerning ultrasonography is required. This paper proposes a CAD by a neural network, which diagnoses a tissue characterization of liver. This system studies three kinds of teacher signal (conditions of liver : normal, chronic hepatitis and liver cirrhosis). As a result, this system is confirmed that it can distinguish between normal of liver and liver cirrhosis. In addition, auto optimization causes the adverse effect for the discriminate accuracy. Our future purpose is the discriminate accuracy improved by increasing the number of teacher signal.
