ClariCT.AI

ClariCT.AI is a software device employing pre-trained deep learning models to denoise and enhance CT images in DICOM format, with the goal of improving image quality, especially for low-radiation dose acquisitions across multiple anatomical regions. Its system involves receiving images from CT scanners, processing them for noise reduction and enhancement using AI, and then transmitting the improved images to PACS, with its efficacy substantiated through comprehensive non-clinical testing against industry standards and diverse clinical datasets.

ClariCT.AI is a software device designed for networking, communication, processing, and enhancement of CT images in DICOM format, compatible with various CT scanner manufacturers and models. Its core purpose is to denoise and enhance CT images, with a particular focus on those acquired using lower radiation doses, applicable to scans of the head, chest, heart, and abdomen. The system operates by receiving DICOM images from CT modalities, applying pre-trained deep learning models for noise reduction and image enhancement, and subsequently transmitting these processed images to a Picture Archiving and Communication System (PACS). This methodology aims to improve the quality of low-dose images, including both Filtered Back Projection (FBP) and Iterative Reconstruction (IR) images, potentially rendering them diagnostically viable. Non-clinical performance testing confirmed ClariCT.AI's adherence to recognized international standards such as ISO 14971 (risk management) and NEMA-PS 3.1-3.20 (DICOM), as well as FDA guidance for software in medical devices and interoperability. Validation was rigorously conducted using an ACR CT Accreditation Phantom and a diverse range of clinical processed datasets. These datasets included paired low and high-dose images from the same patients, IR and FBP images, and data representing various patient demographics (genders, ages, body weights, races, ethnicities) and scan conditions from different CT scanner vendors and for different organs. The testing results demonstrated that ClariCT.AI meets its intended use, predefined functionality requirements, and acceptance criteria, indicating its effectiveness in image enhancement and noise reduction.

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 ClariCT.AI's deep learning model for noise reduction compare quantitatively (e.g., SNR, CNR, radiologist scoring) to the predicate device's regularization process, especially in very low-dose scenarios?; What specific pre-training datasets were used for the deep learning models in ClariCT.AI, and how does the model generalize to CT images from scanners or protocols not heavily represented in the training data?; Are there any known limitations or contraindications for using ClariCT.AI, particularly concerning specific types of artifacts, patient populations, or rare pathologies where the denoising algorithm might obscure subtle findings?; How does ClariCT.AI handle the balance between noise reduction and preservation of fine diagnostic details or textures, and what validation was performed to ensure diagnostically critical information is not lost or distorted?; What are the computational requirements (e.g., processing time per study, GPU specifications) for ClariCT.AI in a typical clinical workflow, and how does it integrate with existing PACS and CT scanner infrastructure?