OBGESS: Automating Original Bender Gestalt Test Based on One Stage Deep Learning

Abstract

Abstract Hand sketch psychological data are mysterious and can be used to detect mental disorders early and prevent them from getting worse and with irreversible consequences. The Original Bender Gestalt Test is a psychology test based on hand-sketched patterns. Mental disorders require an automated scoring system. Unfortunately, there is no automatic scoring system for the Original Bender Gestalt test for adults and children with high accuracy. Automating the Original Bender Gestalt test requires 3 phases: Phase 1, collecting a comprehensive Original Bender Gestalt dataset called OBGET. Phase 2, classifying patterns by a proposed method called MYOLO V5; and Phase 3, scoring classified patterns according to associated rules of psychological standard criteria. This research reviews a comprehensive OBGET dataset that includes 817 samples, labeling samples for mental disorders by a psychologist, statistical analysis, the proposed semi-automatic labeling of patterns, patterns classification applied the proposed modified YOLO V5 called MYOLO V5, and automatic scoring of drawing patterns. MYOLO V5 accuracy is 95% and the accuracy of the proposed method called OBGESS as a mental disorder detection is 90%. In this research, a new automatic computer-aided psychological hand sketch drawing test has been proposed.

Keywords

• OBGESS
• Automatic scoring
• Detecting mental disorders
• Psychological test dataset
• Object detection
• Classification