AI-Rad Companion (Pulmonary)

AI-Rad Companion (Pulmonary) SW version VA20 is an advanced image processing software utilizing machine and deep learning to provide quantitative and qualitative analysis of adult CT DICOM images, thereby supporting clinicians in the assessment and evaluation of lung diseases. Its key functionalities include comprehensive lung and lobe segmentation, detection and follow-up of lung nodules with change quantification, AI-driven pulmonary density assessment with opacity scoring, and offers both cloud and on-premise edge deployment options.

AI-Rad Companion (Pulmonary) is advanced image processing software designed to support radiologists and various physicians (including those in emergency medicine, specialty care, urgent care, and general practice) in the evaluation and assessment of lung diseases using previously acquired Computed Tomography (CT) DICOM images from adult patients. Its core functionalities include the automated segmentation and measurement of the complete lung and individual lung lobes, the identification of areas with lower Hounsfield Unit values (compared to a predefined threshold) potentially indicative of conditions like emphysema, the segmentation and measurement of solid lung nodules with their assignment to corresponding lung lobes, and an interface to the external syngo.CT Lung CAD medical device. The software also facilitates the correlation of segmented lung nodules identified in current scans with known prior images, enabling quantitative assessment of any changes. The enhanced SW version VA20 of AI-Rad Companion (Pulmonary) leverages machine and deep learning algorithms to introduce significant new capabilities. A key update is the lung lesion follow-up assessment feature, which automatically compares currently segmented lung lesions with corresponding priors and quantitatively assesses changes; non-clinical bench testing for this feature demonstrated a sensitivity of 94.3% and an average Positive Predictive Value (PPV) of 99.1% in identifying lesion pairs across 200 cases. Another major addition is the AI-based pulmonary density assessment, which segments opacity regions within the lung, calculates the percentage of these voxels relative to lobe and total lung volume, and assigns a score (0-4 per lobe, summed for a total score of 0-20) based on opacity percentage. Furthermore, the software now offers bi-directional lesion diameter measurements (including short-axis diameter) derived from existing segmentation contours and introduces flexible deployment options, supporting the existing cloud deployment as well as a new on-premise "edge" deployment for local clinical data processing, while the edge system remains connected to the cloud for monitoring and maintenance. The software has been validated for CT data from Siemens Healthineers, GE Healthcare, and Philips scanners using specified reconstruction techniques.

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 AI algorithm for pulmonary density assessment differentiate between various types of opacities (e.g., ground-glass, consolidation, fibrotic changes) and has its accuracy been validated against pathological findings?; What is the minimum detectable change in lesion size, volume, or density for the lung lesion follow-up feature to reliably quantify progression or regression, and how does this compare to manual RECIST criteria?; How does the performance of the lung segmentation and nodule detection algorithms vary across different CT scanner vendors, models, and a wider range of reconstruction kernels than those explicitly listed in the validation?; What are the specific data security and patient privacy measures implemented for the on-premise 'edge' deployment, especially concerning its connectivity to the cloud for monitoring and maintenance?; Are there plans to incorporate other imaging biomarkers, such as airway measurements or vascular changes, into future versions of AI-Rad Companion (Pulmonary) to provide a more comprehensive lung assessment?