Hepatic VCAR

Hepatic VCAR is a CT image analysis software that utilizes deep learning algorithms, specifically convolutional neural networks, to automate the segmentation of the liver and hepatic artery from DICOM CT scans. This technology aims to assist clinicians by providing tools for assessing liver morphology, quantifying lesions, and tracking changes over time through enhanced visualization and measurement capabilities.

Hepatic VCAR is a CT image analysis software package designed for the analysis and visualization of liver CT data obtained from DICOM 3.0 compliant CT scans. Its primary purpose is to assist clinicians in assessing liver morphology, including the identification and measurement of liver lesions that exhibit a different CT appearance from surrounding tissue, and to track changes in these features over time. The software provides automated tools for the segmentation and measurement of the liver, liver lobes, liver segments, and liver lesions. It offers initial 3D segmentation, vessel analysis, visualization, and quantitative analysis of liver anatomy, with functionalities allowing users to adjust contours and confirm final segmentations. The system can be deployed on various platforms, including Advantage Workstation, AW Server, CT Scanners, PACS stations, or cloud environments. A significant technological advancement in this version of Hepatic VCAR is the introduction of two new algorithms utilizing deep learning technology, specifically convolutional neural networks. One algorithm automates the segmentation of the liver, producing an editable liver contour, while the other semi-automates the segmentation of the hepatic artery based on an initial user input point, with the resulting segmentation also being user-editable. This contrasts with previous methods that used traditional deterministic algorithms for liver segmentation and manual tools for vascular structures. The performance of these new deep learning algorithms was validated through bench tests against expert-annotated ground truth, confirming they performed as expected. Furthermore, an assessment involving board-certified radiologists evaluating clinical sample images demonstrated the capability of the deep learning-based liver and hepatic artery segmentation.

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. What specific types or characteristics of liver lesions are most accurately segmented by the deep learning algorithms, and are there known limitations (e.g., size, contrast, proximity to vessels)?; How does the performance (accuracy, speed, robustness) of the deep learning-based hepatic artery segmentation compare to experienced radiologists performing manual segmentation, particularly in complex vascular anatomies?; What was the range of variability in segmentation results observed during the radiologist assessment, and how does the software guide users in adjusting contours for optimal accuracy?; Are there specific CT acquisition protocols or contrast enhancement phases that optimize the performance of Hepatic VCAR's deep learning algorithms for liver and vessel segmentation?; How does the software handle temporal changes in liver morphology and lesion characteristics when comparing studies performed at different time points, especially concerning registration accuracy and consistency of segmentation?