Optimizing Medical Imaging Through Radiation Reduction and Advanced Denoising
This data-driven framework combines a simulation processor and a denoiser model to transform low-dose, noisy medical scans into high-quality images to reduce patients’ exposure to radiation. While medical imaging is crucial for diagnosing and monitoring diseases, the high radiation doses from conventional scans pose significant health risks to patients and limit the frequency of their use. Currently, the standard practice involves high-dose CT scans that, while effective, expose patients to significant levels of radiation. Annually, approximately 1.6 million people in the U.S. face an increased risk of cancer due to the radiation exposure from CT scans, especially those requiring frequent imaging. However, advancements in low-dose CT technology are transforming the market by reducing these risks while still delivering high-quality images. This growing demand for safer diagnostic tools is driving the CT scan market, which was valued at $8.5 billion in 2023 and is projected to reach $12.8 billion by 2030, as healthcare providers prioritize patient safety alongside diagnostic accuracy.
Researchers at the University of Florida have developed the first medical image restoration model to use phantom and deep learning for real low-dose noise simulation and denoising. This framework involves a machine learning strategy that outperforms state-of-the-art denoising algorithms. It significantly reduces radiation exposure by efficiently denoising low-dose CT scans to produce images of comparable quality to high-dose scans and enhances throughput, allowing physicians to evaluate more patients within the same timeframe.
Application
A data-driven framework that enhances low-dose CT scans by restoring high-quality images, reducing radiation exposure, and speeding up diagnostics for safer, more efficient healthcare
Advantages
- Reduces patient radiation exposure by enhancing the quality of low-dose scans, making diagnostic imaging safer
- Saves time in generating high-quality images, allowing healthcare providers to scan and evaluate more patients in a shorter time frame
- Works with any type of medical scan image, offering versatile applications across multiple diagnostic tools
- Outperforms current state-of-the-art denoising algorithms, providing superior restoration of low-quality medical images
- Uses a flexible learning method to restore clear medical images from low-quality scans, making it adaptable for different neural networks
- Combines noise and tissue data, making it easier to reduce noise and enhance image clarity
- Continuously improves image quality by having the noise simulation and denoising process work together, ensuring better results over time
Technology
Cyclic Simulation and Denoising is a framework that produces high-quality images from low-dose medical scans, effectively reducing radiation exposure for patients. It combines a simulator, which extracts low-dose noise and tissue features from different image spaces, with a denoiser that effectively reduces the noise while simultaneously restoring tissue details for clearer, more accurate imaging. The cyclic feedback loop between the two components continuously optimizes learning.
Brochure
