Intelligent Dementia Diagnosis Based On Partial Volume Correction And Images Deep Derivation

Due to the introduction of other pixel information in the nearest neighbor region of pixels in the improvement process of arterial spin labeling image of dementia patients in the traditional region-based partial volume effect improvement method,the improvement results may be excessively fuzzy and the details of brain tissue structure may be seriously missing.In this study,a partial volume effect improvement method based on single pixel information is proposed.Without introducing additional information of other pixels,partial volume effect improvement is only carried out based on the information of single pixel,which avoids the defects of traditional region-based improvement method.The deep learning model was used to synthesize the arterial spinal labeling images from structural magnetic resonance images.Since the arterial spinal labeling images are a kind of functional magnetic resonance images,the synthesis process is a heterogeneous synthesis from structural magnetic resonance images to functional magnetic resonance images.The proposed pixel-wise method is used to improve the partial volume effect of the synthesized arterial spin labeling images.Then,combining the structural magnetic resonance images with the synthesized arterial spin labeling images,multiple diagnostic tools are used to intelligently diagnose the condition of dementia patients.