Hand vein recognition using Mask R-CNN

Medical imaging technologies play a crucial role in critical applications such as dorsal hand vein detection. Traditional methods for locating veins can pose challenges in treatment and diagnosis, particularly for children, elderly individuals, obese patients, and individuals with darker skin tones. In this study, an automatic dorsal hand vein detection system was developed using a deep learning-based Mask R-CNN model for dorsal hand vein segmentation. A dataset provided by Boğaziçi University, consisting of 1,575 dorsal hand vein images collected from 100 individuals, was utilized. To enhance vein structures, contrast-enhancing image processing techniques were applied, and the vein networks were manually annotated in detail. The Mask R-CNN model was then trained on this dataset to perform segmentation, and its accuracy was optimized. The study achieved an accuracy rate of 75.3%. The obtained results demonstrate the potential of this method to provide a practical and reliable solution in clinical settings..