(207 days)
uAI Portal is a software solution intended to be used for viewing, manipulation, and storage of medical images. It supports interpretation of examinations within healthcare institutions. It has the following additional indications:
The Lower Extremity Vessel Analysis is intended to provide a tool for viewing, manipulating, and evaluating CTA images of lower extremities.
The Head and Neck Vessel Analysis is intended to provide a tool for viewing, manipulating, and evaluating imaging datasets acquired with head and neck CTA.
The Coronary Analysis is intended to provide a tool for viewing, manipulating imaging datasets acquired with CCTA.
The Pulmonary Artery Analysis is intended to provide a tool for viewing, manipulating, and evaluating imaging datasets acquired with CTPA.
The Aorta Analysis is intended to provide a tool for viewing, and evaluating imaging datasets acquired with aorta CTA.
uAI Portal is a comprehensive software solution designed to process, review and analyze CT studies. It can transfer images in DICOM 3.0 format over a medical imaging network or local file system. These images can be functional data, as well as anatomical datasets. It can be at one or more time-points or include one or more time-frames. Multiple display formats including VR, MIP, MPR, Probe, CPR, and SCPR. A trained, licensed physician can interpret these displayed images as well as the statistics as per standard practice.
uAI Portal contains the following applications:
- The Lower Extremity Vessel Analysis
- . The Head and Neck Vessel Analysis
- . The Coronary Analysis
- . The Pulmonary Artery Analysis
- . The Aorta Analysis
Here's a breakdown of the acceptance criteria and study details for the uAI Portal device, based on the provided FDA 510(k) summary:
1. Acceptance Criteria and Reported Device Performance
The device's performance was evaluated using the Dice coefficient, which measures the similarity between the algorithm's segmentation result (P) and the ground truth (G). The formula used is: $DICE = \frac{2 * G \cap P}{G + P}$.
| Application | Algorithm | Acceptance Criteria (Dice) | Reported Average Dice |
|---|---|---|---|
| Coronary Artery | Vessels segmentation | 0.85 | 0.920 |
| Heart segmentation | 0.90 | 0.980 | |
| Head and Neck Vessel | Head vessel segmentation | 0.85 | 0.902 |
| Neck vessel segmentation | 0.90 | 0.967 | |
| Aorta | Trunk segmentation | 0.90 | 0.946 |
| Branches segmentation | 0.80 | 0.846 | |
| Pulmonary Artery | Arteries segmentation | 0.85 | 0.953 |
| Veins segmentation | 0.85 | 0.933 | |
| Lower Extremity Artery | Arteries segmentation | 0.80 | 0.892 |
All reported average Dice values exceed their respective acceptance criteria.
2. Sample Size and Data Provenance
- Test Set Sample Size: 150 images.
- Data Provenance for Test Set: Collected from the US. The data covered a variety of demographics (gender, age), equipment (SIEMENS, GE, TOSHIBA), and image characteristics (with/without artifacts, with/without anatomical variation).
- Data Provenance for Training Set: Images collected from China. The data set ensured a variety of data for different gender, age, equipment, and CT protocol.
- Retrospective/Prospective: Not explicitly stated, but the mention of "images collected from US" and "images collected from China" for testing and training datasets respectively, combined with the ground truth establishment process involving expert annotation of existing images, strongly suggests a retrospective study design.
3. Number of Experts and Qualifications for Ground Truth
- Number of Experts:
- Initial Annotation: Two (2) Chinese radiologists.
- Adjudication: One (1) American Board-Certified Radiology adjudicator.
- Qualifications of Experts:
- Initial Annotation: Each radiologist had at least 5 years of clinical experience. They were hospital employees and independent of United Imaging.
- Adjudication: The adjudicator was an American Board-Certified Radiology adjudicator with at least 10 years of clinical experience.
4. Adjudication Method for the Test Set
- Method: A 2+1 adjudication method was used.
- Two (2) Chinese radiologists independently annotated the vessel mask for each patient case, resulting in two sets of annotations.
- An American Board-Certified Radiology adjudicator (the "1" in 2+1) reviewed both sets of segmented images.
- Based on their assessment, the adjudicator selected the most accurate segmentation set as the final ground truth. If necessary, they would make modifications until a satisfactory ground truth was established.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- No, a MRMC comparative effectiveness study was NOT done. The study focused on the standalone algorithmic performance (Dice coefficient) against an expert-established ground truth. There is no mention of human readers assisting or comparing performance with and without AI.
6. Standalone (Algorithm Only) Performance
- Yes, a standalone (algorithm only without human-in-the-loop performance) study was done. The entire "Performance Verification" section details the algorithm's performance in segmenting various anatomical structures by comparing its output (P) to the established ground truth (G) using the Dice coefficient.
7. Type of Ground Truth Used
- The ground truth used was expert consensus / adjudicated expert annotation. Specifically, two radiologists initially annotated cases, and a third, more experienced radiologist adjudicated and finalized the ground truth.
8. Sample Size for the Training Set
- The exact sample size for the training set is not specified, only that images were "collected from China".
9. How the Ground Truth for the Training Set was Established
- The document states that "Algorithm training of uAI Portal software has been conducted on images collected from China as training dataset." However, it does not explicitly detail how the ground truth for this training dataset was established. It implies that these images had associated ground truth data for the algorithm to learn from, but the process of creating that ground truth for the training set is not described in the provided text.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).