AVIEW Lung Nodule CAD

AVIEW Lung Nodule CAD is a computer-aided detection software designed to improve radiologists' accuracy and efficiency in identifying pulmonary nodules on chest CT scans from asymptomatic individuals. It employs a Deep Convolutional Neural Network algorithm to automatically detect nodules between 3 and 20mm, functioning as an adjunctive second reader and integrating with existing radiology workflows.

AVIEW Lung Nodule CAD is a software tool designed to assist radiologists in detecting pulmonary nodules, ranging from 3 to 20mm in diameter, in chest CT images of asymptomatic individuals. The system employs a Deep Convolutional Neural Network (CNN) based algorithm, trained on chest CT data, to automatically identify these nodules. It operates in compliance with DICOM standards, allowing integration with Picture Archiving and Communication Systems (PACS), and provides a dedicated user interface for analyzing, identifying, storing, and transmitting quantitative data about the detected nodules. Intended as an adjunctive tool, it functions as a second reader after the radiologist's initial review, allowing clinicians to confirm or dismiss the CAD's findings. The device's performance was notably validated using non-contrast CT images, primarily acquired from Siemens SOMATOM series scanners, and it is designed to be used with the cleared AVIEW platform. The clinical efficacy of AVIEW Lung Nodule CAD was assessed through a multi-reader, multi-case retrospective study involving 151 chest CTs and eleven board-certified radiologists. This study demonstrated statistically significant improvements when radiologists used the AI tool: the area under the curve (AUC) for nodule detection increased from 0.73 (unaided) to 0.92 (aided), and sensitivity rose from 0.68 to 0.91. Furthermore, the rate of false positives per scan decreased from 0.48 to 0.28, and reading time was reduced with AI assistance. A separate standalone performance evaluation on 282 datasets yielded an AUC of 0.961, sensitivity of 0.907, and specificity of 0.704, with a sensitivity of 0.889 at a false positive rate of 0.504 per scan, indicating robust standalone algorithmic performance.

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 AVIEW Lung Nodule CAD's performance compare on CT images acquired with different reconstruction kernels (smooth, medium, sharp) beyond the listed typical examples, and is there an optimal kernel setting for maximal detection efficacy?; The documentation states performance validation on Siemens SOMATOM CT scanners. What are the specific performance metrics or expected variability when using images from other CT manufacturers or models, particularly those with different noise profiles or image reconstruction algorithms?; Can the system differentiate between various types of pulmonary nodules (e.g., solid, subsolid, ground-glass opacity) within the 3-20mm range, and does its sensitivity or specificity vary significantly across these nodule types?; The reader study showed a decrease in false positives per scan with AI assistance. What are the common characteristics of false positives generated by the AI, and how does the system address potential confounders like scarring or vascular structures?; How does the integration with the AVIEW platform enhance the workflow beyond nodule detection, and are there plans to incorporate nodule characterization or risk stratification features directly within the CAD module?