Structural alterations anisotropy-based measured for different areas for the most common types of dementia diseases could be a biomarker of brain impairment. The current work aims to assess whether texture anisotropy can discriminate both healthy versus Alzheimer’s and Pick’s patients based on regional evaluation while maintaining high predictive power. The investigated area is reduced from the whole-brain surface to three major lobes (i.e., frontal, temporal and parietal). A predictive model was proposed to associate a disease with a specific area in the brain based on the anisotropy values. Simultaneous analysis of 1680 measurements from 105 brain magnetic resonance images acquired as T2w and PD sequences was performed to establish the significance of the model. The cerebral calcinosis disease has been used as artificial ground truth. The association based on textural anisotropy between targeted diseases and control patients was performed by using Pearson’s correlation coefficients. A new proposed consistency index investigated the texture anisotropy relevance for all image’s types and all analyzed classes and regions. The validation study is based on area under the receiver-operating characteristic curve that depicted the overall diagnostic performance of the texture anisotropy in each region. The proposed model demonstrated that texture anisotropy is accurate solution in diagnosis of Alzheimer’s and Pick’s diseases when the investigated area is reduced to major lobes, with sensitivity >90% and specificity >80%.
相似文献