DEVELOPING STEM CELL CLINICAL TRIALS IN SPINAL CORD INJURY THROUGH ARTIFICIAL INTELLIGENCE TECHNIQUES

Abstract

Primary characteristics that facilitate the treatment of SCI with MSCs include their proliferative potential, immunomodulation capacity, and differentiation ability. Nevertheless, traditional approaches employed to determine stem cell viability and readiness for therapeutic use are vague, detrimental, or prolonged. In this study, we present a new image analysis method for predicting viability and therapeutic potential of mesenchymal stem cells in spinal cord injury models through their detection and classification. The algorithm was developed based on phase contrast microscopy images acquired during the primary and early logarithm stages of MCS expansion concerning the specific treatment protocols designed for spine cord injury. Stem cells were identified through edge detection, thresholding, and morphological operations. In order to resolve inter-cells within clusters, H-minima transform, and Hidden Markov Models (HMM) were applied. Marker-controlled watershed techniques were used to segment out the clustered cells in order to obtain single-cell data. Consequently, morphometric and textural features were extracted, and machine learning techniques were employed in order to classify on the basis of morphological phenotypes of MSCs. The algorithm was tested externally over 899 MSCs selected from 596 culture images of spinal cord injuries. The model has proven to be 97% sensitive and 88% specific, producing 98% and 98% precision for detection and segmentation of MSCs, respectively per image. For classifying MSC phenotypes expanding during early and mid-lag phases, the AUC values were 0.99 (CI95 = 0.976–0.988) as obtained in the charts. The presented approach attributes a high level of accuracy and reliability in segmenting and classifying MSCs according to their shapes. This approach is useful for non-invasively determining the quality of MSC culture and can help in the maintenance of quality control based on the morphology of the cells in relation to the development of treatment regimens using mesenchymal stem cells for repair of damaged spinal cord tissues.