| Summary: | Peripheral intravenous (IV) access is an important issue in daily practice in hospital as it is a common practice in the medical field. However, it will be a difficult task if the dorsalhand veins are not clear or obvious for intravenous access. The poor visibility of dorsal vein may result in wrong puncturing which causes patient to suffer from pain and even will lead to permanent damage of vein. Hence, a number of imaging methods have been implemented to expose the veins. At present, there are limited literature regarding the specific study to justify the enhancement algorithm which can perform effectively for dorsal hand vein imaging. Therefore, this research is set-up to develop a modified image enhancement
algorithm for dorsal hand vein. Firstly, the grayscale hand vein image obtained from NIR imaging with noise undergoes grayscale enhancement by applying Contrast Limited
Adaptive Histogram Equalization (CLAHE). Then, the adaptive thresholding method is implemented on the filtered grayscale image for vein pattern segmentation purpose. After
image segmentation, the input image is converted to binary image. The noisy binary vein pattern is then enhanced using a combination of Feed-Forward Neural Network (FFNN),
Area Opening (AO) and Binary Median Filter (BMF). Finally, the enhanced image is evaluated by examining the image’s sensitivity, specificity and accuracy of the enhanced
image through comparison with the ground truth images. The evaluation results between modified image enhancement algorithm are compared with the existed algorithm. The evaluation results shows that the AO-FFNN-BMF sequence produces the highest sensitivity, specificity and accuracy for both input images. The proposed technique has produced the clearest vein patterns in terms of connectivity and smoothness than the other binary
enhancement techniques.
|
|---|