(227 days)
UW LV Analysis Software assists in measuring left ventricular volume at end diastole and end systole, ejection fraction, and regional wall motion around the ventricular contour. The intended applications of the software are: 1) to assist in evaluating a patient with suspected heart disease, 2) to measure the effectiveness of therapy, 3) to assess risk or prognosis, and 4) in clinical trials evaluating new therapies.
The method comprises five main parts: a) reviewing the digital images, b) selecting the image frames to analyze, c) identifying and tracing the border of the left ventricle, d) measuring left ventricular volume from the traced border and calculating ejection fraction, and e) measuring regional wall motion. The UN LV Analysis Software allows the user to review a series of images in digital format, and manually select frames for analysis. Identifying and tracing the ventricular border is performed manually using the Of the techniques developed to measure left LV Analysis Software. UW ventricular volume from a traced border, the area length method is generally the most accurate, and is the method used by the UW LV Analysis accepted as Software. Of the techniques for measuring left ventricular wall motion, the centerline method has been proven useful for clinical trials, and is the method used by the UW LV Analysis Software. Calibration to correct for image magnification is required for volume measurement, but not for calculation of the ejection fraction or for analysis of regional wall motion.
Here's an analysis of the provided text, focusing on the acceptance criteria and the study details, presented in the requested format:
Acceptance Criteria and Device Performance Study
The provided document describes the UW LV Analysis Software, which assists in measuring left ventricular volume, ejection fraction, and regional wall motion. This software is largely a manual tool with automated calculations based on user input.
1. Table of Acceptance Criteria and Reported Device Performance
| Metric | Acceptance Criteria | Reported Device Performance |
|---|---|---|
| Difference from reference programs (volume & regional wall motion) | < 2.5% | Less than 2.5% (observed during licensing/installation and internal upgrades) |
| Volume measurement variability | Not explicitly stated as acceptance criteria, but reported as a characteristic. | 8 ml |
| Ejection Fraction variability | Not explicitly stated as acceptance criteria, but reported as a characteristic. | 0.04 |
| Regional Wall Motion (anterior wall) variability | Not explicitly stated as acceptance criteria, but reported as a characteristic. | 0.24 SD/chord |
| Regional Wall Motion (inferior wall) variability | Not explicitly stated as acceptance criteria, but reported as a characteristic. | 0.33 SD/chord |
2. Sample Size Used for the Test Set and Data Provenance
The document does not explicitly define a "test set" in the context of a formal validation study with a distinct sample size for evaluating acceptance criteria. Instead, it describes:
- During software upgrades: "the data of thousands of patients were analyzed by both the new and previous versions and compared". This suggests a large, retrospective set of patient data, likely from the University of Washington's clinical archives.
- During licensing/installation: The standard of accuracy was applied "when licensed to other companies and installed on their systems". This implies a comparison against the University of Washington's programs, possibly using a smaller, but representative, dataset for each installation.
Data Provenance: The data primarily originated from the University of Washington. The nature of the data (retrospective/prospective) isn't explicitly stated for all uses, but the "data of thousands of patients" used for upgrades strongly suggests retrospective analysis of existing clinical data.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts
The concept of "experts establishing ground truth for the test set" as a distinct activity is not described. The software relies on manual tracing by users/clinicians. The "standard of accuracy" during licensing/upgrades was comparison against the "programs at the University of Washington." This implies that the University of Washington's version, and the measurements derived from it by experienced users, served as the de facto reference.
While not explicitly stated, it's reasonable to infer that the "programs at the University of Washington" were used and validated by experienced cardiologists or technicians over 13 years, given their use in NIH-sponsored trials. However, the specific number or qualifications of these "experts" for creating a ground truth dataset are not provided.
4. Adjudication Method for the Test Set
No formal adjudication method is described for generating a ground truth test set. The validation seems to be based on:
- Comparison between software versions: The new version's results were compared to the preceding version's results.
- Comparison against the "University of Washington" standard: Installed programs were compared against the UW programs.
This implies a direct comparison rather than a consensus-based adjudication process.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done, If So, What Was the Effect Size of How Much Human Readers Improve with AI vs. without AI Assistance
No MRMC comparative effectiveness study is described. The software is presented as an assistance tool for human readers (users manually trace borders). There is no mention of a study comparing human performance with and without the software, nor any effect size regarding human improvement. The software's primary role is to provide a standardized tool and calculations for manual input, rather than an AI-driven interpretation tool.
6. If a Standalone (i.e., algorithm only without human-in-the-loop performance) Was Done
No standalone algorithm performance study was done. The UW LV Analysis Software description clearly states: "Identifying and tracing the ventricular border is performed manually using the UW LV Analysis Software." This explicitly indicates a human-in-the-loop process for the critical step of border identification. The software then performs calculations based on these manual tracings.
7. The Type of Ground Truth Used
The ground truth used for validation (during upgrades and licensing) was essentially based on measurements derived from previous, trusted versions of the same software, operated by experienced users at the University of Washington. In essence, it's a form of expert consensus/established clinical practice as embodied by the long-standing use of the UW programs. It is not pathology, or direct outcomes data.
8. The Sample Size for the Training Set
The document does not describe a distinct "training set" in the context of machine learning. The software is described as a rule-based system ("area length method," "centerline method") that performs calculations based on manual input. It is not a supervised learning model that requires a labeled training set in the typical AI sense.
9. How the Ground Truth for the Training Set Was Established
Since there is no training set mentioned in the context of machine learning, the question of how its ground truth was established is not applicable. The software's methods (area length, centerline) are established clinical methodologies, not learned from data.
{0}------------------------------------------------
JAN 17 1907
Application K962153
1962153
8462828
Summary of Safety and Effectiveness Information
UW LV Analysis Software assists in measuring left ventricular volume at end diastole and end systole, ejection fraction, and regional wall motion around the ventricular contour. The method comprises five main parts: a) reviewing the digital images, b) selecting the image frames to analyze, c) identifying and tracing the border of the left ventricle, d) measuring left ventricular volume from the traced border and calculating ejection fraction, and e) measuring regional wall motion. The UN LV Analysis Software allows the user to review a series of images in digital format, and manually select frames for analysis. Identifying and tracing the ventricular border is performed manually using the Of the techniques developed to measure left LV Analysis Software. UW ventricular volume from a traced border, the area length method is generally the most accurate, and is the method used by the UW LV Analysis accepted as Software. Of the techniques for measuring left ventricular wall motion, the centerline method has been proven useful for clinical trials, and is the method used by the UW LV Analysis Software. Calibration to correct for image magnification is required for volume measurement, but not for calculation of the ejection fraction or for analysis of regional wall motion.
There is no substantial hazard of death or injury to the patient associated with using this software. The intended applications of the software are: 1) to assist in evaluating a patient with suspected heart disease, 2) to measure the effectiveness of therapy, 3) to assess risk or prognosis, and 4) in clinical trials evaluating new therapies.
Although it is believed that no hazard to the patient exists this from software, considerable effort has been directed to analyzing the sources of error or variability in the measurements, and to developing methods of avoiding minimizing them. During software development, the author frequently or exercised all functions to assure proper operation. After upgrades of the software to improve performance, the data of thousands of patients were analyzed by both the new and previous versions and compared, to document that the revision had not introduced new errors or inconsistencies, or significantly altered the measurement results.
The standard of accuracy applied to validation of this software when licensed to other companies and installed on their systems, was to demonstrate the installed programs differed from the programs at the University of Washington by less than 2.5%. In internal testing of upgrade versions at the University of Washington, the measurements of volume and regional wall motion differed from those of the preceding version by less than this. The measurement variability is 8 ml for volume, 0.04 for ejection fraction, and 0.24 and 0.33 SD/chord for motion in the anterior and inferior walls, respectively.
Versions of these programs were used over the past 13 years at the University of Washington. They were also used to make endpoint measurements for a number of NIH-sponsored and agency-sponsored clinical trials. The selection of the UW LV Analysis Software programs for the NIH's TIMI trial is an indication of NIH's endorsement of their performance and accuracy.
§ 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).