A Hybrid Approach to Endoscopic Image Enhancement for Better Disease Detection

Abstract

Endoscopy is a critical tool in diagnosing gastrointestinal (GI) tract diseases, such as ulcers, cancers, and inflammatory conditions. However, the quality of endoscopic images often varies due to factors like inadequate lighting, image noise, and limited resolution, which can impede accurate diagnosis. Image enhancement techniques offer promising solutions to improve image clarity, sharpness, and contrast, thereby aiding in the early detection of GI disorders. The diagnostic performance of endoscopic procedures can be compromised by low-quality images, which may result in missed lesions, inaccurate assessment of disease progression, or unnecessary biopsies. In particular, traditional endoscopic imaging is often limited by suboptimal resolution, image distortion, and noise interference. There is a need for robust enhancement algorithms that can address these challenges and facilitate better disease detection. We propose a hybrid approach combining multiple image enhancement algorithms, including contrast enhancement, noise reduction, and resolution improvement, to optimize the quality of endoscopic images. Specifically, we used a combination of histogram equalization for contrast enhancement, wavelet-based denoising for noise reduction, and super-resolution algorithms to upscale image resolution. The algorithms were applied to a dataset of 500 anonymized gastrointestinal endoscopic images, with objective quality metrics like Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and subjective assessment through expert evaluation. Our proposed enhancement pipeline demonstrated a significant improvement in image quality compared to the original endoscopic images. On average, the PSNR improved by 12.3 dB, and the SSIM increased by 0.18, indicating better structural preservation and contrast. Expert evaluations confirmed a 35% increase in the visibility of critical features, such as lesions and mucosal abnormalities. Furthermore, the enhanced images facilitated more accurate disease diagnosis, with a 20% improvement in the detection rate of early-stage cancers and ulcers.