Dysarthria is a neurological speech disorder that severely impacts communication. The goal of this review is to assess recent advances in dysarthria speech detection. Our study systematically examined dysarthria datasets, preprocessing techniques, and feature extraction methods. A number of key findings emerged, including (1) a trend toward multimodal and self-supervised learning, (2) the emergence of transfer learning techniques, and (3) improved classification accuracy for dysarthria subtypes. In generalizing across diverse datasets, severe dysarthria cases remain challenging. A comprehensive dataset and robust methods for detecting severe dysarthria are needed, according to our analysis. Our proposal integrates advanced audio processing with machine learning techniques. Researchers and practitioners can benefit from the review, which provides valuable insight into dysarthria treatment.