True Enhance DL

True Enhance DL employs deep learning algorithms to post-process standard 120 kVp single-energy CT images, aiming to improve image contrast by estimating 50 keV monochromatic GSI-like characteristics. This software provides non-quantitative, adjunctive visual information to aid clinicians, with specific models for different vascular enhancement phases, and is validated for use across all age groups.

True Enhance DL is a deep learning-based image processing software from GE Healthcare Japan Corporation, designed to operate with the Revolution Ascend family of CT systems. Its core function is to enhance the contrast of 120 kVp single-energy computed tomography images of the body. The technology employs deep learning algorithms trained to estimate monochromatic, 50 keV Gemstone Spectral Imaging (GSI)-like images, thereby aiming to improve visualization from standard acquisitions. It offers four user-selectable deep learning models tailored for different contrast enhancement phases: CT Angiography, Arterial, Portal/Venous, and Delayed. True Enhance DL provides non-quantitative, adjunctive information, which must be interpreted alongside the original 120 kVp images, and is applicable for patients of all ages on non-GSI Revolution Ascend systems. The efficacy and safety of True Enhance DL were assessed through comprehensive non-clinical and clinical evaluations. Non-clinical testing included software verification and validation, risk management, and image quality performance assessments using standardized phantoms and metrics, including IEC 61223-3-5 Ed. 2. Clinical validation was conducted via a reader study involving four board-certified radiologists who evaluated retrospectively collected clinical cases, comparing images processed by True Enhance DL against the original 120 kVp images for image quality and contrast improvement. Additionally, CT number characteristics in various anatomical regions were studied by comparing True Enhance DL outputs with 50 keV monochromatic and 120 kVp reference images from a GSI system. These studies confirmed that True Enhance DL improves image contrast relative to the original single-energy 120 kVp images.

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 True Enhance DL's estimated 50 keV GSI-like images compare to actual hardware-generated 50 keV GSI images in terms of specific diagnostic tasks, beyond general contrast improvement?; What are the computational requirements (e.g., processing time) for True Enhance DL, and how might this impact clinical workflow in high-throughput environments?; Are there specific patient populations or pathological conditions where True Enhance DL provides a more significant diagnostic advantage over standard 120 kVp images?; How robust are the four deep learning models (CT Angiography, Arterial, Portal/Venous, Delayed) to variations in contrast media injection protocols or patient-specific physiological differences?; Given that True Enhance DL provides non-quantitative information, what is the inter-observer variability among radiologists in interpreting the enhanced contrast and its clinical significance?