Ai-Driven Prediction Models For Medical Image Enhancement And Analysis Using Deep Learning

Abstract

Thyroid diseases, including cancer, demand for accurate medical imaging for diagnosis and treatment direction. Important MRI imaging of the thyroid occasionally suffers with poor resolution and noise. Conventional imaging enhancement techniques may not be able to effectively solve noise or capture small features required for a successful diagnosis, therefore lowering the image quality and diagnostic accuracy. For feature extraction and classification, we provide an artificial intelligence prediction model on multi-scale deep convolutional neural network (CNN). Our method reduces noise and solves resolution enhancement, hence improving thyroid MRI images. For this work we used a bespoke CNN architecture and a 500 thyroid MRI image collection. Not only considerably outperforming current techniques, our model generated a Structural Similarity Index (SSIM) of 0.89 and a Peak Signal-to---- Noise Ratio (PSNR) of 32.5 dB. The improvements brought diagnosis accuracy 15% above more traditional techniques.