Detection of CBF deficits in neuropsychiatric disorders by an expert system: a 99Tcm-HMPAO brain SPET study using automated image registration.

The aims of this study were to develop an objective method for assessing rCBF deficits using a statistical image analysis protocol and to validate its effective use in clinical practice. 99Tcm-HMPAO brain SPET images were acquired for 40 normal subjects, 10 patients with Alzheimer's disease and 10 patients with depression. Automated image registration was used to standardize the size and shape of the brain structures for all subjects. The images of the first 30 normal subjects were used to construct a normal database. The CBF images of the other 10 normal subjects and the 20 patients were compared voxel by voxel with the normal database to map CBF abnormalities by statistical evaluation. The results were compared with the clinical reports of CBF images. The expert system detected all rCBF deficits reported by the nuclear physicians. Some additional areas with special information, like atrophy and bilateral asymmetry, were also identified by the expert system. We conclude that this expert system can delineate CBF deficits with sufficiently high accuracy, differentiating normal from abnormal CBF images using voxel-based comparisons. The use of an expert system improves rCBF SPET image evaluation.