A hybrid image enhancement based brain MRI images classification technique

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Accurate classification of brain Magnetic Resonance Imaging (MRI) images aids radiologists by identifying normal versus abnormal scans.
• Statistical analysis methods are common for MRI classification but are sensitive to image quality.

Purpose of the Study:

• To test the hypothesis that improved image quality during pre-processing enhances the performance of statistical classification approaches for brain MRI.
• To develop and validate an enhanced image pre-processing technique for improved medical image analysis.

Main Methods:

• Implemented an improved image enhancement technique including median filtering for noise removal and histogram equalization for contrast enhancement.
• Extracted features using discrete wavelet transform and reduced them using color moments (mean, standard deviation, skewness).
• Trained a deep neural network (DNN) for classifying brain MRI images as normal or pathological.

Main Results:

• The proposed method achieved a classification accuracy of 95.8%.
• This accuracy significantly surpasses previous state-of-the-art techniques in brain MRI classification.
• The results validate the critical role of image enhancement in improving classification performance.

Conclusions:

• Enhanced image pre-processing is crucial for improving the accuracy of automated medical image classification.
• The developed technique demonstrates potential for enhancing various medical image analysis tasks.
• This approach offers a promising avenue for reducing radiologist workload and improving diagnostic efficiency.