Eclipse II with Smart Noise Cancellation

The Eclipse II with Smart Noise Cancellation software utilizes a Convolutional Neural Network to intelligently reduce noise in digital X-ray images, aiming to enhance diagnostic quality. This technology enables significant radiation dose reduction for patients while maintaining or improving image clarity for both adult and pediatric radiography, excluding mammography.

The Eclipse II with Smart Noise Cancellation is a software module designed to enhance digital radiographic images from x-ray devices, applicable to both adult and pediatric patients, excluding mammography. Integrated within the ImageView application, its primary function is to improve image quality by reducing noise in projection radiography acquisitions from computed radiography (CR) and digital radiography (DR) detectors, specifically supporting the Carestream DRX family. The core of the Smart Noise Cancellation (SNC) feature is a Convolutional Neural Network (CNN), based on a U-Net architecture. This neural network is trained using clinical images deliberately degraded with simulated noise to effectively identify and map a "Noise Field" in acquired images, enabling targeted noise reduction prior to standard Eclipse image processing. The methodology involves passing a preprocessed image through the trained CNN, which estimates and allows for the suppression of image noise. This enhancement aims to improve image attributes such as brightness, latitude, contrast, detail, sharpness, and noise appearance, thereby assisting clinicians in diagnosis. Clinical testing, involving a concurrence study with board-certified radiologists, evaluated images using visual difference preference and RadLex scales. The results demonstrated that the software can achieve significant radiation dose reduction—up to 50% for Cesium Iodide (CsI) panel images and 40% for Gadolinium Oxysulfide (GOS) panel images—while maintaining or even improving diagnostic image quality compared to images acquired at nominal doses, across various examination types, detector technologies, and exposure levels.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the Smart Noise Cancellation CNN vary across different anatomical regions and patient demographics (e.g., pediatric vs. adult, varying body habitus)?; What are the specific quantitative metrics, beyond visual preference and RadLex scales, used to validate the "Noise Field" estimation accuracy by the U-Net architecture?; Are there any known limitations or specific image artifacts that might be introduced or exacerbated by the SNC module, particularly at the maximum claimed dose reduction levels?; How adaptable is the pre-trained CNN model to new detector technologies or imaging protocols not included in the initial training dataset, and what is the process for retraining or fine-tuning?; What is the computational overhead (e.g., processing time increase) associated with the SNC module, and how does it impact overall workflow efficiency in a clinical setting?