(142 days)
iCardio.ai EchoMeasure is software that is used to process previously acquired DICOM-compliant cardiac ultrasound images, and to make measurements on these images in order to provide automated estimation of several cardiac measurements. The data produced by this software is intended to be used to support qualified cardiologists, sonographers, or other licensed professional healthcare practitioners for clinical decision-decision-making.
iCardio.ai EchoMeasure is indicated for use in adult patients.
iCardio.ai EchoMeasure is a software device used to process previously acquired DICOM-compliant transthoracic cardiac ultrasound images. The software provides automated view classification and quality check of images to then provide several automated estimation of cardiac anatomical measurements and quantities.
iCardio.ai EchoMeasure is a comprehensive software application that seamlessly integrates image pre-processing and quality check of standard cardiac ultrasound views and provides automated measurements of standard cardiac parameters and measurements.
iCardio.ai EchoMeasure is designed to sort through and determine the eligibility criteria for downstream processing, including image quality, and appropriate cardiac view. The following pre-processing steps are considered in making a determination about image eligibility for processing:
- Echocardiographic view classification
- Echocardiographic view overall image quality -
- -End-Diastolic and End-Systolic frame identification
iCardio.ai EchoMeasure automatically sorts through and recognizes these key parameters to then allow an image to pass for automated processing for measurement of several cardiac parameters, including:
-
- Left Ventricular Volume (A2C, A4C, and Biplane; Systole and Diastole)
-
- Left Ventricular Diameter (Systole and Diastole)
-
- Right Ventricular Diameter
-
- Posterior Wall Thickness
-
- Aortic Annulus Diameter
-
- Left Ventricular Outflow Tract Diameter
-
- Sinus of Valsalva Diameter
-
- Sinotubular Junction Diameter
-
- Left Atrium Dimension
-
- Interventricular Septal Thickness
Machine learning based view detection, quality grading, key frame selection, automated keypoint detection and segmentation form the basis of the software's automated analysis.
iCardio.ai EchoMeasure output is intended for consumption by 3rd party software and hardware vendors. Additionally iCardio.ai has a native browser interface for reviewing the report summary as well as a functionality to download the available report in PDF format. The iCardio.ai EchoMeasure browser interface allows the end user to view both 2D image and cine loops determined by the software and to review the automated measurements produced. It is the option of the reviewing clinician to accept, reject, edit, or ignore the output provided by iCardio.ai EchoMeasure.
A report, automatically generated from the calculated parameters, is returned to the interpreting clinician. This software device aims to aid diagnostic review and analysis of echocardiographic data, patient record management, and reporting. It also features tools for organizing and displaying quantitative data from cardiovascular images acquired from ultrasound scanners. It is exclusively for use by qualified clinicians.
Here's an analysis of the acceptance criteria and study detailed in the provided text:
Acceptance Criteria and Device Performance
1. Table of Acceptance Criteria & Reported Device Performance
The acceptance criteria for iCardio.ai EchoMeasure's performance were based on the Bi-variate Linear Regression Coefficient Slope (BLRSC). The device was designed to estimate the "worst-case" error, defined as the difference between the software output and the mean of three clinician-derived annotations. The acceptance criterion was that the estimated worst-case BLRSC (based on the 95% CI) for each endpoint must be above a certain predetermined threshold. The study's conclusion explicitly states that "In no instance did the worst-case BLRSC for a given measurement (calculated based on the 95% confidence interval) fall below the predetermined, minimum allowable BLRSC threshold."
| Measurement | Metric | Acceptance Criteria (Implicit) | Reported Device Performance (Value [95% CI] BLRSC) |
|---|---|---|---|
| Aortic Annulus Diameter | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.952 [0.829, 1.082] |
| Left Ventricular Outflow Tract Diameter | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 1.112 [0.970, 1.255] |
| Sinus of Valsalva Diameter | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.932 [0.848, 1.015] |
| Sinotubular Junction Diameter | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.773 [0.676, 0.869] |
| Left Atrial Diameter | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.888 [0.830, 0.944] |
| Left Ventricular Diameter (Systole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.860 [0.776, 0.945] |
| Left Ventricular Diameter (Diastole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.791 [0.710, 0.869] |
| Right Ventricular Diameter (Diastole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.786 [0.715, 0.854] |
| Interventricular Septal Thickness | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.833 [0.731, 0.934] |
| Posterior Thickness | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.785 [0.664, 0.904] |
| Left Ventricular Volume (A4C-Systole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 1.059 [0.977, 1.158] |
| Left Ventricular Volume (A4C-Diastole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.943 [0.869, 1.013] |
| Left Ventricular Volume (A2C-Systole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.936 [0.777, 1.048] |
| Left Ventricular Volume (A2C-Diastole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 1.005 [0.917, 1.096] |
| Biplane LV Volume (Systole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.906 [0.795, 0.993] |
| Biplane LV Volume (Diastole) | BLRSC | Worst-case BLRSC (lower bound of 95% CI) above a predetermined minimum allowable threshold. | 0.972 [0.893, 1.054] |
2. Sample Size for Test Set and Data Provenance
- Sample Size: 200 comprehensive echocardiography studies from 200 distinct patients. A single DICOM was selected for each relevant view (PLAX, A2C, or A4C).
- Data Provenance: Retrospective, sampled from two independent clinical sites from two different US states. This was done to assure a wide sample of imaging data and patient demographics. No data from these sites was used for the training or tuning of the algorithm.
3. Number of Experts and Qualifications for Ground Truth (Test Set)
- Number of Experts: Three (3)
- Qualifications: Experienced US-based cardiac sonographers.
4. Adjudication Method for Test Set
The ground truth was established using the mean of three (3) clinician-derived annotations per case. This implies a consensus-based approach or averaging of independent expert measurements.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
The provided text does not mention a multi-reader multi-case (MRMC) comparative effectiveness study to assess the effect of AI assistance on human reader performance. The study described is a standalone performance study.
6. Standalone Performance Study
Yes, a standalone performance study was conducted. The objective was to demonstrate successful device performance using prospectively-defined success criteria for each endpoint, specifically evaluating the "worst-case" error for linear and volumetric measurements against clinician-derived ground truth.
7. Type of Ground Truth Used
The ground truth used was expert consensus based on manual measurements and segmentations performed by experienced clinicians (the mean of three experienced US-based cardiac sonographers).
8. Sample Size for Training Set
The text does not specify the sample size for the training set. It only mentions that the sonographers used for the standalone study were independent of those used to annotate the training data, and that data from the two clinical sites used for the test set was not used for training or tuning.
9. How Ground Truth for Training Set was Established
The text does not explicitly detail how the ground truth for the training set was established, other than noting that different sonographers were involved compared to the test set ground truth establishment.
§ 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).