Intelligent IoMT Sonography System for Accurate Thyroid Nodule Detection

This study presents a method for preprocessing and classifying ultrasound images using deep learning. The preprocessing steps include resizing, intensity clipping, pixel normalization, noise reduction with bilateral filtering, contrast enhancement via CLAHE, image sharpening, morphological dilation, and edge detection with Canny. These techniques improve the performance of machine learning models, particularly for small or complex datasets [1][2]. The processed images are classified using a ResNet50-based Convolutional Neural Network (CNN), fine-tuned with a thyroid ultrasound dataset. The ResNet50 model, pre-trained on ImageNet and adapted for this task, achieves high accuracy, demonstrating the effectiveness of both the preprocessing steps and the transfer learning approach in medical image classification [1][2]. Keywords - Ultrasound Imaging, Medical Image Processing,CLAHE,Convolution Neural Networks(CNNs),ResNet 50,Deep Learning Classification,Thyroid Imaging.