Intelligent Hand Vein Image Exposure System To Aid Peripheral Intravenous Access

• Main Author: Yakno, Marlina
• Format: Thesis
• Language: English
• Published: 2013
• Subjects: TK7800-8360 Electronics
• Online Access: http://eprints.usm.my/49041/1/Marlina%20Binti%20Yakno_HJ.pdf; http://eprints.usm.my/49041/

Difficulty in achieving peripheral intravenous (IV) access in some patients is a clinical problem. These difficulties may lead to some negative impacts such as fainting, hematoma and pain associated with multiples punctures. As a result, ultrasound and infrared imaging devices have been used to aid IV access. Although these devices have shown to be able to aid IV ac- cess, infrared system has not been able to produce satisfactorily clear vein patterns and using ultrasound device is time consuming. Therefore, this research focuses on developing a hand vein exposure system with enhanced image of hand vein patterns to aid IV access. It consists of three major sub-systems namely, a hand vein image-acquisition system, image processing component and hand vein image-projection system. The image acquisition system consists of forty eight near-infrared light emitting diode with wavelength of 0.89mm. The image process- ing system involves six stages. In the first stage, a noisy hand vein image is filtered using a feed-forward neural network (FFNN) based on standard median filter. In the second stage, a newly proposed technique based on finger-webs and finger-tips characteristics is applied to ob- tain a larger region of interest (ROI). In the third stage, the ROI images are enhanced using a combination of fuzzy histogram hyperbolization and contrast limited adaptive histogram equal- ization. Then, in the fourth stage, vein patterns are segmented using local adaptive threshold. In the fifth stage, a noisy binary vein patterns are enhanced using a combination of FFNN pixel correction, binary median filter and massive noise removal. In the last stage, an enhanced vein patterns are registered into the original hand vein layout. Finally, the last sub-system projects the registered vein patterns onto a patient’s hand. A combination of FFNN pixel correction, binary median filter and massive noise removal as proposed has been able to increase the sensitivity of binary image of vein patterns up to 10.664% from the original binary image. The average difference in standard deviation between the enhanced images and their truth image is 0.02016. This difference is the smallest in comparison to images obtained based on existing image enhancement methods. The developed system has shown to be able to enhance hand vein image patterns for easy IV access. It has the potential to significantly reduce the average IV access time and most importantly, it could shed patients’ fear towards IV access.