Izotropic's Breakthrough AI to Transform Breast CT Imaging

Izotropic Corporation has introduced a new proprietary artificial intelligence algorithm into its IzoView Breast CT Imaging System, aiming to raise global standards for image quality and patient safety in breast cancer screening.

The company’s machine-learning reconstruction algorithm, developed with The Johns Hopkins University School of Medicine, is trained on 15 years of breast CT data. It improves image quality by reducing noise without increasing radiation exposure, addressing one of the biggest challenges in CT imaging.

Image noise, the grainy texture often seen in CT scans, increases as radiation doses are lowered. Since minimising dose is crucial for patient safety, there has always been a compromise between clarity and safety. Traditional denoising techniques, including Model-Based Iterative Reconstruction (MBIR) and Deep Machine-Learning Reconstruction (DMLR), have helped but remain limited in routine use. MBIR is slow and resource-heavy, while most DMLR methods struggle with the specific challenges of breast CT, often compromising anatomical detail.

Izotropic’s algorithm uses a novel self-supervised deep learning approach that processes raw X-ray data before reconstruction. This method delivers faster results, preserves natural breast tissue texture, avoids the need for paired high- and low-dose training data, and is suited to high-throughput clinical environments.

Protected as a trade secret, the algorithm provides Izotropic with a durable competitive edge, as it is specifically trained on highly specialised breast CT data. This positions IzoView as a strong contender to set new global benchmarks for low-dose, high-quality breast imaging.

The development also supports future advances in computer-aided diagnostics. By generating cleaner, high-resolution images that prioritise patient safety, IzoView creates ideal datasets for AI diagnostic tools, reinforcing Izotropic’s role in shaping the future of imaging and precision healthcare.