Medical Image Super-resolution Algorithm Using Feature Based Regression Analysis

Project Abstract / Summary : Medical Image super-resolution (MISR) algorithms focus on increasing the resolution of medical images by establishing a relationship between corresponding low and high resolution images. Many state-of-the-art medical image enhancement algorithms are not capable of handling real-time data and processing or need large amount of training data to provide acceptable results. We derive a linear relationship using redundant information across the different scales of the medical image itself along with a pre-trained operator mapping by extracting priori information offline from a training data-set which is faster to process and use bootstrapping approach as a back-propagation method for iterative noise reduction and quality enhancement. Matrix based regression methods have shown superior results when applied to this setting, as they use matrix level image information. In this paper, we extend the concept of a matrix operator to a low level feature super-resolution algorithm that performs edge enhancement and de-noising best suited for medical images as they are costly and generally of poor quality. Although it is competitively effective in resolving edges and other low level features which characterize X-Ray images, MRI scans etc; it is computationally less efficient when compared to image based operators. We propose a novel patch based implementation, in which the decision to implement the costly feature regression is based on the feature components present in that particular patch. Smooth and gradual patches are resolved by image operators and patches with strong features are resolved by feature operators. The proposed algorithm achieves competitive performance efficiency and is verified by experimental results for X-Ray images, MRI scans and CT scan images.

Why did you choose to work on this project topic : The diagnoses of many fatal and life crippling diseases lie in the resolution of the medical images. The resolution of these images direct the prognoses of the medical analysis and is therefore highly important. Also state-of-the-art medical assistance is costly and is not real-time. Thus we want to bring the prognoses in a smart-phone, with real time cheap medical prognoses assisted by our super-resolution algorithm, that feeds input to a classifier which predicts the disease. Needless to say, the resolution of the images, directly affect the classifier output and the prognoses and thus we present this algorithm.

Project Highlights : This project is scalable and implementable in real-time. As it already is producing state-of-the-art results, setting up an app or a software platform and implementing the prognoses part of the medical analysis will easily be done quickly with the right funding. Also research needs to be done in the field of medical imaging for it is as important as inventing new drugs. With more funding in this domain, daily check up, predictive prognoses and follow up in a smartphone is on course.


Project Category : Electrical / Electronics / Communication
------------------------------------------------------
Institute/College Name: SSN College of Engineering
City: Chennai
State: Tamilnadu
Participating Team From: Final Year