(17 days)
Medilog Darwin Holter Analysis is indicated for the replay and analysis of ECG data prerecorded on Medilog series ambulatory recorders.
It is indicated for use in the analysis, display, editing and report generation of ambulatory ECG data as part of the assessment of cardiac rhythm disturbance and myocardial ischaemia. Typically it is intended for patients requiring analysis of 24 hour ambulatory ECG recordings as determined by a medical practitioner.
Medilog Darwin Holter Analysis is designed to acquire, analyse and store data recorded on Medilog series ambulatory ECG recorders. Editing facilities are offered to allow some of the information to be modified. The processed data can be printed in the form of an analysis report. The device is not a life supporting, implanted or otherwise critical device, and it is does not provide on-line warnings or alarms.
Medilog Darwin Holter Analysis is designed to operate on an industry standard PC. Customers purchase the device as a software kit for installation on their own compatible PC.
Here's a breakdown of the acceptance criteria and study details for the Medilog Darwin Holter Analysis device, based on the provided 510(k) summary:
1. Table of Acceptance Criteria and Reported Device Performance
The 510(k) summary does not explicitly state numerical acceptance criteria (e.g., "accuracy must be >90%"). Instead, the primary acceptance criterion appears to be demonstrating substantial equivalence to the predicate device, specifically by performing as well as or better than the predicate device, and compliance with relevant ANSI/AAMI standards.
| Acceptance Criterion (Inferred from 510(k)) | Reported Device Performance |
|---|---|
| Compliance with ANSI/AAMI EC38:1998 (Ambulatory electrocardiographs) | Exceeds predicate device performance on AHA, MIT, ESC, and NST databases. |
| Compliance with ANSI/AAMI EC57:1998 (Testing and reporting performance results of cardiac rhythm and ST segment measurement algorithms) | Exceeds predicate device performance on AHA, MIT, ESC, and NST databases. |
| Performance in identifying isolated VEBs, couplets, bigeminy, pauses, bradycardia, tachycardia in non-paced recordings. | Closer to manual count in more cases than the predicate device. |
| Performance in identifying abnormalities in paced recordings. | Closer to manual count in more cases than the predicate device. |
| Overall safety and effectiveness compared to predicate device. | As safe and effective, and performs as well as or better than the predicate device. |
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size: Not explicitly stated as a single number.
- For the standard databases: The document mentions "AHA, MIT, ESC and NST databases." These are well-established public databases commonly used for Holter analysis algorithm testing. The exact number of recordings within these databases is not provided in this summary, but they typically contain numerous subjects/recordings.
- For parallel tests: "non-paced three-channel Holter recordings from different customers" and "paced three-channel Holter recordings from different customers." The specific number of customers or recordings is not provided.
- Data Provenance:
- AHA, MIT, ESC, NST databases: These are international, publicly available and widely recognized databases for ECG analysis. Their original provenance (e.g., country of origin, retrospective/prospective collection) is inherent to the specific database (e.g., MIT-BIH Arrhythmia Database, European ST-T Database). The 510(k) does not specify the exact records used, but they are generally retrospective.
- Customer data for parallel tests: "from different customers." The country of origin is not specified, nor whether it was retrospective or prospective, though general "customer recordings" often imply retrospective collected data.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts
- Number of Experts: Not specified.
- Qualifications of Experts: Not specified. The ground truth for the "manual counts" is established by human review, but the expertise of those performing the manual counts is not detailed. For the standard databases (AHA, MIT, etc.), the ground truth is typically meticulously annotated by multiple cardiologists or experts, but this document does not provide details on the specific qualification of those experts.
4. Adjudication Method for the Test Set
- For the "manual counts" used in the parallel tests against customer data, the adjudication method is not specified.
- For the standard databases (AHA, MIT, ESC, NST), the ground truth labels are typically established through a rigorous consensus process by multiple experts, but the specific adjudication method (e.g., 2+1, 3+1) for these databases, nor for the manual counts in this study, is detailed in the 510(k).
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, an MRMC comparative effectiveness study was not explicitly done in the sense of comparing human readers with AI assistance versus without AI assistance.
- The study focuses on the standalone performance of the algorithm and its comparison to the predicate device and manual counts. The device's role is described as providing "additional information" to a qualified physician who "reviews and interprets" the reports. This suggests an assistance tool, but not a study designed to quantify human performance improvement with the tool.
6. If a Standalone (i.e. algorithm only without human-in-the-loop performance) Was Done
- Yes, a standalone algorithm-only performance evaluation was conducted.
- The summary explicitly states: "The performance of the candidate device was tested on the AHA, MIT, ESC and NST databases..." and "The unedited analysis results generated by each device were then compared against manual counts..." This indicates that the algorithm's raw output, without human intervention or editing, was directly evaluated against established ground truth.
7. The Type of Ground Truth Used
- Expert Consensus / Reference Databases: For the AHA, MIT, ESC, and NST databases, the ground truth is established through expert-annotated reference labels.
- Manual Counts: For the parallel tests on customer data (both paced and non-paced), the ground truth was established by "manual counts of isolated VEBs, couplets, bigeminy, pauses, bradycardia, tachycardia, etc." This implies human expert review and counting to establish the true presence and number of these events.
8. The Sample Size for the Training Set
- Not specified. The 510(k) summary only describes the testing of the device and its algorithm, not the development or training phase.
9. How the Ground Truth for the Training Set Was Established
- Not specified. Similar to the training set size, the 510(k) does not provide details on how the training data was curated or how its ground truth was established. This information is typically part of the development documentation but not always included in the 510(k) summary focused on validation.
§ 870.1425 Programmable diagnostic computer.
(a)
Identification. A programmable diagnostic computer is a device that can be programmed to compute various physiologic or blood flow parameters based on the output from one or more electrodes, transducers, or measuring devices; this device includes any associated commercially supplied programs.(b)
Classification. Class II (performance standards).