Evaluation of Preprocessing Techniques for Brain Analysis Using Compressed and Uncompressed Magnetic Resonance Imaging

Abstract

Digital Image Processing (DIP) contributes with many advantages to medical diagnostics. Images can be optimized by improving their quality, which allows specialists to better locate tissue damages, or other anomalies. The planning and execution of surgeries, design of prosthesis, monitoring and evaluation of progression of diseases can greatly benefit from DIP. In particular, Magnetic Resonance Imaging (MRI) is a commonly used technique for medical diagnostics. It is highly accepted for its high precision and resolution in anatomical explorations, allowing an accurate analysis of interest area and the behavior of surrounding tissues. The image sequences captured in different three-dimensional (3D) planes by MRI analysis, allow specialists to determine better diagnoses and treatment for patients. However, the data loss caused by commonly used image compression formats could affect results, if the images are used without preprocessing techniques. This paper will evaluate differences between compressed and uncompressed images, proposing a methodology to improve the quality of compressed images recovering information by the uses of bi-linear and bi-cubic interpolations. The obtained results will be measured with Signal-Noise Ratio (SNR) and variance differences for each case to validate the preprocessing techniques applied. According to the obtained results, data lost by compression algorithms could be recovered by the proposed interpolations techniques.