Development of Alzheimer's Disease Recognition using Semiautomatic Analysis of Statistical Parameters based on Frequency Characteristics of Medical Images

Parallel nonlinear analysis of weighted brain's gray and white matter images for Alzheimer's dementia diagnosis

“…The interesting result accomplished showed a 93 percent accuracy which was done just by a linear discriminator. However, they analyzed all 260 features and didn't eliminate potentially dependent data, and that's what we considered in our previous works [2], [3], and also we do it in the current work. We eliminate redundant data in this project by applying Principal Component Analysis (PCA) on the T1 and T2 images in order to reduce dependent data and time needed for executing the proposed texture analysis method.…”

“…In this regard, we notice that T1 images have information about gray and white matters of brain, while T2 images have information about gray matters. Also we notice that AD defects mainly gray and white matters of brain more than black and marginal parts of the brain [3]; therefore, for this specific disease, we believe that T1 images have more data than T2 images. Consequently, we should define the neighborhood in T1 images with more accuracy and details compared to T2 images.…”

“…Basically, registration includes two stages of interpolation: pixel position interpolation and gray-level interpolation. For performing interpolation process, the well-known radial basis function r 2 log r 2 is used which has been commonly used in medical images registration [3]. It should be noted that the affine terms (which could be added to r 2 log r 2 ) have not been used in the proposed algorithm, because the results (without using the affine terms) were already satisfactory, therefore adding a new term, actually adds an unnecessarily cost to the registration optimization cost function by increasing the computational time and complexity of the system structure.…”

“…For this purpose, initially landmarks need to be accurately placed on each image; however, because of the above-mentioned reason, landmarks selection for each image might be different from others. The landmarks are located and placed on those points of the brain which are geometrically and anatomically more important [3]. The number of landmarks, according to the accuracy criterion, has been selected to be 10.…”

“…Based on our previous publications, we believe that the tissue analysis can be a reliable tool to diagnose various diseases. For example, in addition to two projects for diagnosing AD by tissue analysis [2], [3], we also utilized this method for diagnosis of Multiple Sclerosis (MS) [4]. Additionally, we could diagnose various breast diseases with the help of texture analysis [5]- [7].…”