Mori H, Suzuki Y, Hisada K, Kojima K, Tonami N
Invest Radiol. 1975 May-Jun;10(3):251-7. doi: 10.1097/00004424-197505000-00010.
Using 240 true positive brain scans, a computer system for the differential diagnosis of brain lesions has been evaluated. Eighty-six parameters were extracted from brain scan findings without relationship to neurological signs and symptoms, and the likelihood method was adopted as an example of mathematical logic. The results of our experiment indicated that the overall accuracy was 77 per cent for the maximum likelihood method. The digital computer gave satisfactory results, particularly for diseases such as infarct, meningioma, acoustic neurinoma, and subdural hematoma. In spite of several problems to be solved, this method could provide invaluable help in differential diagnosis of brain lesions.
利用240例脑扫描确诊病例，对一种用于脑损伤鉴别诊断的计算机系统进行了评估。从脑扫描结果中提取了86个与神经体征和症状无关的参数，并采用似然法作为数理逻辑的一个实例。我们的实验结果表明，最大似然法的总体准确率为77%。数字计算机给出了令人满意的结果，尤其是对于诸如梗死、脑膜瘤、听神经瘤和硬膜下血肿等疾病。尽管有几个问题有待解决，但这种方法可为脑损伤的鉴别诊断提供宝贵的帮助。
