n Technology Platform
"Medical Image Processing Method for Blood Vessel based on Image Fusion method" is for separating the vascular regions from nonvascular ones by removing the noise component and amplifying the signal of vascular region on the blood vessel image. The proposed method using Artificial Neural Network is a novel fusion method on the medical image processing. Futhermore, the proposed method can make the enhanced blood vessel image in order to correctly diagnose disease. While CT(Computed Tomography) and MRI(Magnetic Resonance Imaging) use a contrast medium for obtaining the blood vessel image, the proposed method use the vascular image captured by NIR (Near-Infrared) LED and camera. Therefore, the proposed method has the advantages compared to existing CT and MRI.
n Background and unmet needs
A lot of people are suffered for the diseases related to blood vessel such as Buerger's disease, Raynaud's phenomenon and rheumatoid disease. Until now, these diseases has been diagnosed by using MRI and CT. However, MRI and CT have the problems of high cost, inconvenience to patient and the side effects by using a contrast medium. To solve these problems, the blood vessel images captured by NIR (Near Infrared) LED (Light Emitting Diode) and camera are used in this invention. However, as the captured blood vessel image is blurred by skin scattering, the discrimination of vascular and nonvascular areas decreases. In order to correctly diagnose diseases, the medical image processing methods, which enhance low quality image to high quality image, are actively developed in medical field.
To obtain high quality image, Kim introduced an image merging method for two high dynamic range (HDR) images of different exposure. After two HDR images under different exposure time were captured, two HDR images were combined by weights computed from brightness and chroma component of HDR images. Also, Gaussian function was used for preventing the sparkle noises to obtain the enhanced image. However, when capturing two HDR images on the blood vessel, rotation and translation are generated by user's movement in the captured two blood vessel images. The rotation and translation in two images are difficult to be estimated and the procedure of image alignment from two HDR images takes much processing time. So, since the distortion of image can be generated by wrong alignment of two images, this method is improper to apply for obtaining the vein image of good quality.
Shi et al. proposed the method of extracting hand vein patterns from low-quality images. They used the matched filter, wiener filter and average filter to remove the noise and enhance the visibility quality of blood vessel image. However, as the result image was blurred by using many image filters, the discrimination of vascular areas from nonvascular ones in blood vessel image was degraded.
n Discovery and Achievements
The proposed method ("medical image processing method for blood vessel based on image fusion method") can enhance not only the quality of blood vessel image to discriminate the vascular areas and nonvascular ones, but offer enhanced blood vessel images based on image fusion methods, by which medical doctor can accurately diagnose a disease.
The proposed method uses adaptive Gabor filtering, edge detection, MLP and score level fusion method. Adaptive Gabor filtering method is used for amplifying the signal of vascular region on the blood vessel image. To separate the sections between vascular and nonvascular regions, edge detection method is used. The MLP and score level fusion methods are used for enhancing the quality of blood vessel image by combining two images (blood vessel image and the edge image) to one image (blood vessel image of enhanced quality).