Computer aided diagnosis of myocardial bull's eye map for detection of coronary artery lesion

运动和再分布TL-201心肌SPECT虽然是检测冠状动脉病变的有效方法，但衰减和康普顿散射会使心肌图像出现伪影，阻碍正确诊断。因此，TL-201心肌SPECT诊断的准确性取决于图像解译人员的技能。为了更客观和正确地诊断，用牛眼显示器进行量化分析是有用的。此外，由正常患者标准化的牛眼显示范围和严重程度图比简单的牛眼图更有助于客观和正确的诊断。这些范围和严重程度图的解释是对缺陷模式的简单识别，可能适合于计算机分析。目前为止，牛眼图是根据心肌短轴SPECT的定量周向分析制作的，排除了心尖面积的信息。为了建立更准确的包含心尖区的牛眼图，我们研制了用于定量分析的放射状长轴SPECT，它是由短轴SPECT重建的。牛眼图的正常模式有性别差异，我们使用了男女两种正常模式。对于临床冠状动脉正常、8种病变类型(LAD、RCA、LCX、LAD RCA、LAD LCX、RCA LCX、LAD RCA LCX病变)的临床患者在有限的时间内收集非常困难，对神经计算机的初步学习还不够充分。神经计算机的诊断准确率高于经验不足的医生，但不如熟练的医生。因此，通过大量的患者数据对神经计算机进行更多的训练，将提高神经计算机的诊断能力，并在一段时间内成为一种很好的计算机辅助诊断设备。

Although exercise and redistribution Tl-201 myocardial SPECT is useful method for detection of coronary artery lesion, attenuation and Compton scatter make artifacts on myocardial image and obstruct correct diagnosis. Therefore, diagnostic accuracy of Tl-201 myocardial SPECT depends upon the skill of image interpreter. For more objective and correct diagnosis, quatintative analysis by Bull's eye display is usefull. Moreover, extent and severity map of Bull's eye display standardized by normal patients are more useful for objective and correct diagnosis than simple Bull's eye map. The interpretation of these extent and severity map is simple recognition of defect pattern and may be suitable for computer analysis.So far, Bull's eye map was made from quantitative circumferential analysis of myocardial short-axial SPECT,and the information of apical area was excluded. For creation of more correct Bull's eye map including apical area, we developed radial long axial SPECT for quantitative analysis, which was reconstracted from short axial SPECT.Normal pattern of Bull's eye map had gender difference, and we used 2 kinds of normal patterns for men and women. The initial learning for neurocomputer was not enough, because collection of clinical patients having normal coronary artery and 8 kinds of abnormal coronary lesion patterns (LAD,RCA,LCX,LAD+RCA,LAD+LCX,RCA+LCX,LAD+RCA+LCX lesions) was very difficult for limited periods. The diagnostic accuracy of neurocomputer was superior to interpretation by inexperienced docters, but inferior to that by skillful docters. Thus, the more training of neuro-computer by many patient's data will improve the diagnostic ability of neuro-computer and become a good computer aided diagnostic device in some time.