One of the vital areas within medical systems analysis is the benchmark dataset. Currently, many related research fields are using AI methods for more accurate and quicker diagnosis, prognosis, and generalization in many areas such as neurology, cardiology, retinal image processing, and so on. To make a decision, we need accurate data to support our analysis. Consequently, we need to consider performance and decide more efficiently. There are many neurological disorders, ranging from neurodegeneration to trauma, or other related pathologies, in which all of these illnesses have their own sub-type disorders. However, we will treat a patient, and most treatment will be either cognitive or medicinal therapy after we diagnose the disorders. The fundamental operations for disease analysis are image labeling, preprocessing, and feature extraction from a given dataset. In addition, after we have the dataset for each type of related disease, we will apply image regeneration-based techniques to learn the main principal components, reduce overfitting, and possibly improve the classification methods. We provide detailed benchmarks for subsequent analysis, namely diagnosis and prognosis for fatty liver classification, diabetic retinopathy, different neurological disorders, various mood disorders, cardiac-related treatment courses, and initiating early strokes and survival analysis. The paper is structured as follows: we briefly provide an overview of related data and methodologies for each dataset. The paper presents results of a series of benchmarks of different datasets in medical systems. The dataset is obtained from different sources and medical image data based on different diseases like strokes, diabetic retinopathy, and also based on spatial domain image data for neurological behavior diseases. It offers a performance analysis including precision, recall, F1-score, and the AUC of a machine learning paradigm. The result indicates a general trend in most of the datasets.