A lightweight attention-based deep learning facial recognition system for multiple genetic syndromes

In this paper, we propose a novel facial image-based genetic disorder detection system designed to classify genetic disorders by analyzing distinctive facial deformations caused by various genetic conditions. While state-of-the-art deep learning paradigms have been extensively applied in many fields, their use for genetic disorder detection has not been thoroughly explored in the literature, leaving significant room for advancements. Our proposed framework addresses this gap by introducing a robust and innovative model that enhances data pre-processing, weight optimization, and feature attention within the deep learning architecture. Our system's deep learning-based pre-processing steps are particularly effective at ensuring accurate performance, even when working with low-quality images captured in noisy, non-laboratory-controlled environments. To further improve disorder classification, we have developed an end-to-end framework integrating transfer learning with a deep convolutional neural network and a specialized attention mechanism. This combination allows our system to effectively detect genetic disorders from facial images by emphasizing critical features, leading to superior performance compared to existing state-of-the-art models. Experimental results on a custom-built image dataset demonstrate the significant performance gains achieved by our approach.