SpotLight/SpotLight Duo (with DLIR option)

The SpotLight / SpotLight Duo (with DLIR option) is a dual-tube, multi-slice computed tomography system that generates cross-sectional body images, with a focus on cardiovascular and thoracic applications, by reconstructing X-ray transmission data via a modified FDK cone beam algorithm. Its key innovation is the optional Deep Learning Image Reconstruction (DLIR) algorithm, an AI-based software sub-system designed to improve image quality by reducing noise without loss of diagnostic information, having demonstrated non-inferiority to the ASIR-CV algorithm in performance tests for patients over two years old.

The SpotLight / SpotLight Duo (with DLIR option) is a multi-slice (192 detector rows), dual-tube Computed Tomography (CT) X-ray system designed to produce cross-sectional images of the body, including the head, cardiovascular, and thoracic anatomies. It utilizes computer reconstruction of X-ray transmission projection data acquired from multiple angles as an X-ray source and detector array rotate around the patient. The system can perform Axial, Cine, and Cardiac scans, with or without contrast, and is intended for patients of all ages, although its Deep Learning Image Reconstruction (DLIR) feature is limited to patients above two years old. The primary innovation is the integration of the DLIR algorithm, an AI-based software sub-system, as an alternative image reconstruction method to the existing ASIR-CV noise reduction algorithm, specifically applicable for 250mm Field of View (FOV) and Extended Field of View (EFOV) configurations; this device aims to aid in the diagnosis of diseases or abnormalities and for planning therapy procedures by providing detailed anatomical images. The scientific basis of the SpotLight / SpotLight Duo system involves X-ray attenuation principles, where X-rays pass through the body and their varying attenuation is measured by a detector array to create projection data. This data is then processed using a modified Filtered Back Projection (FDK) cone beam algorithm adapted for dual-tube geometry, with options for noise reduction including the Adaptive Statistical Iterative Reconstruction (ASIR-CV) or the novel Deep Learning Image Reconstruction (DLIR). The DLIR, a key technological advancement, is a deep learning-based algorithm trained to enhance image quality by reducing noise. Non-clinical performance testing on phantoms demonstrated that DLIR effectively decreases pixel-wise noise magnitude without compromising image features, Hounsfield Unit (HU) accuracy, or high-contrast spatial resolution, performing comparably to ASIR-CV. Clinical testing involved a retrospective, blinded evaluation of clinical cases by five certified CT readers, who found that images reconstructed with DLIR provided diagnostic image quality that was not inferior to those reconstructed with ASIR-CV across various anatomies, scan types, and patient BMIs.

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 DLIR algorithm vary across different patient body mass indexes (BMIs) and specific cardiovascular or thoracic pathologies compared to ASIR-CV in terms of subtle lesion detectability?; What are the specific computational requirements (e.g., processing time, memory usage) for the DLIR algorithm compared to ASIR-CV, and how might this impact clinical workflow efficiency?; Are there specific scan protocols or anatomical regions where DLIR demonstrates a more significant improvement in image quality or diagnostic confidence over ASIR-CV, beyond general noise reduction?; What were the characteristics of the training dataset used for the DLIR algorithm (e.g., size, diversity of cases, image acquisition parameters), and how might this influence its generalizability to patient populations or CT scanner hardware not extensively represented in the training?; Beyond non-inferiority, were there any specific image quality metrics (e.g., signal-to-noise ratio, contrast-to-noise ratio, specific artifact reduction) where DLIR showed a statistically significant superiority over ASIR-CV in either phantom or clinical studies?