Project Title: Real-Time Automated Chromosome Segmentation and Probabilistic Inference Platform for Genetic Syndromes Using Artificial Intelligence
Description:
It is crucial to analyze human chromosomes correctly in cytogenetic studies, as well as in evolutionary, embryological and individualized therapeutic planning. Genetic syndromes, malignancies and multi-system abnormalities are linked to structural and numerical abnormalities of chromosomes, which is a form of rapid screening that must be done for early clinical decision making. In order to alleviate the burden and delay of manual karyotyping, this study aimed to create a decision support platform, called Chromosome Scan, that aids in automated chromosome classification and clinical reporting using AI. The evaluation was done using 28,500 human epithelial cell chromosome images from public repositories, standardized, and split into a training, validation, and testing set. Three types of image representations, namely, grayscale image (ChCNg), heatmap image (ChCNh), and RGB image (ChCNr) were included into the proposed Chromosome Classification Networks (ChCN). Furthermore, Grad-CAM was used to enable localized visual explanations, and make the diagnosis more interpretable. In addition, the web-based platform incorporates a genomic reference database which allows retrieval of gene associated information for known chromosomal abnormalities, as well as case management and automated report generation. The proposed framework was validated using experimental results with high validation accuracy of 94.04% for ChCNh, 94.69% for ChCNr, and 95.23% for ChCNg, which showed the effectiveness of the proposed framework for automated cytogenetic analysis.
Uploaded At: 2026-07-02 10:03:51