Impact of Advanced Image Preprocessing Techniques on Automated Chromosome Metaphase Detection with Deep Learning Models

This paper presents a comprehensive analysis of various preprocessing techniques for chromosome metaphase image detection using deep learning models. The primary objective is to enhance the accuracy and efficiency of chromosome detection by optimizing preprocessing steps. Through extensive experimentation, the combination of CLAHE, Gamma Correction, and Median Filtering is shown to outperform other preprocessing methods based on both subjective and objective evaluation metrics. The proposed method achieves exceptional results with an MSE of 3.49, a PSNR of 32.02, an SSIM of 0.99, an Entropy of 4.32, and an Edge Density of 0.03. In terms of detection performance, the method excels with an IoU of 0.90, a precision of 0.95, and an F1-score of 0.92. Moreover, post-detection accuracy is significantly improved, increasing from 73% to 95% when the proposed preprocessing approach is applied before the detection task. This study highlights the effectiveness of the proposed preprocessing pipeline, showcasing its potential for improving automated chromosome detection systems and enabling more accurate analysis of metaphase images. Furthermore, this work can be extended by exploring deep learning-based denoising techniques like GANs, optimizing YOLOv2 through hyperparameter tuning, and evaluating advanced image detection models for improved accuracy and robustness.