Automated schizophrenia detection model using blood sample scattergram images and local binary pattern

The main goal of this paper is to advance the feld of automated Schizophrenia (SZ) detection methods by presenting a pioneering feature engineering technique that achieves high classifcation accuracy while maintaining low time complexity. Furthermore, we introduce a novel data type known as scattergram images, which can be obtained through a simple blood test. These scattergram images provide a cost-efective approach for SZ detection. The scattergram image datasets used in this research consist of images collected from 202 participants, with 106 individuals diagnosed with SZ and the remaining 96 individuals serving as control subjects. Our objective is to assess the ability of scattergram images to detect SZ. To achieve accurate classifcation with minimal computational burden, we propose a feature engineering model based on the local binary pattern (LBP) technique. Initially, a preprocessing method is applied to separate blood cells from the scattergram images, followed by image rotation to ensure robust results. Both 1D-LBP and 2D-LBP are utilized to extract informative features. Our feature engineering model incorporates iterative neighborhood component analysis (INCA) to select the most relevant features. In the classifcation phase, shallow classifers are employed to demonstrate the capability of the extracted features for classifcation. Information fusion is accomplished using iterative hard majority voting (IHMV) to select the most accurate result. We have tested our proposal on the collected two scattergram image datasets and our proposal attained 89.29% and 90.58% classification accuracies on the used datasets, respectively. The findings of this study demonstrate the potential of scattergram images as an effective tool for SZ detection, thus serving as a promising new biomarker in the field. Our auto-detection model of SZ disease is clinically ready for use in hospital settings and outpatient clinics as an additional means to assist clinicians in their diagnostics procedure.