%0 Journal Article
%T Time and Cost Efficient Vein Pattern Recognition and Injection Point Suggestion using Machine Vision Technology
%A Zakria Qadir
%J -
%D 2019
%X Vein detection is an important task for venepuncture. Excessive venepuncture can cause significant problems during an emergency situation due to the lack of experienced medical staff. In this paper, time and cost-efficient portable vein pattern recognition system is proposed. A vein camera mobile application is used to obtain a Near-Infrared Light (NIR) image and then a set of machine vision algorithms are applied to the NIR image in order to find a vein pattern and injection point. Infrared Radiation (IR) sensitive camera can be utilized to produce a Near-Infrared light (NIR) in a specified wavelength range. This camera roughly creates a brightness difference between the vein region and surrounding tissues in the captured image. The vein regions appear to be darker in comparison to surrounding tissues in the images. Past studies depict that cost, time and portability are the main challenges faced during the implementation of this type of camera systems. These challenges can be overcome by using a vein camera mobile application to take an infrared image instead of designing and using an expensive IR camera. The quality of the images captured by a vein camera application is almost the same as the quality of the images captured by an expensive IR camera. The captured image is processed by a sequence of operations such as median filtering, Contrast-limited adaptive histogram equalization (CLAHE) operation, adaptive thresholding, morphological operations, perimeter extraction, and distance transform to determine the vein region, and suggests a location for injection. In particular, CLAHE is the key operation that is employed for contrast enhancement. Although there are techniques to handle vein pattern detection problem, this is the first time a machine vision algorithm including the CLAHE operation is applied to Near-Infrared Light (NIR) images for vein pattern recognition. The proposed algorithm is also capable of injection point suggestion which is very important for venepuncture. Our approach is implemented in MATLAB software and can be applied to both fair and dark skin tones. Evaluations with 21 participants with varying skin tones (fair and dark) show that the proposed approach is especially effective for detecting vein patterns at the back of the hand (with 95.24% accuracy) and wrist (with 76.19% accuracy)
%K Damar £¿r¨¹nt¨¹ tan£¿ma
%K Maliyet ve zaman a£¿£¿s£¿ndan verimli
%K Makine g£¿r¨¹£¿¨¹
%K NIR
%K MATLAB
%K Venopunktur
%U http://dergipark.org.tr/ejeas/issue/47043/570864