Search Results

Indications for use of the Argus Cognitive ReVISION Software are quantification and reporting of right ventricle (RV) structures and function of patients (adults only) with suspected disease to support physicians' diagnosis.

The Argus Cognitive ReVISION Software ("ReVISION Software") assesses global and segmental function of the right ventricle (RV) of the heart via the decomposition of longitudinal, radial, and anteroposterior motions of the RV walls and quantifies their relative contribution to global RV ejection fraction along with longitudinal, circumferential and area strains using 3D datasets obtained by echocardiography. Segmentation is performed on the end-diastolic 3D RV model using a rule-based, non-AI/ML algorithm to have 15 RV segments. This segmentation is then projected to all models in the corresponding cardiac cycle.

Here's a summary of the acceptance criteria and the study proving the device meets them, based on the provided text:

1. Table of Acceptance Criteria and Reported Device Performance

2. Sample Sizes Used for the Test Set and Data Provenance

• Accuracy Validation: 30 subjects (10 healthy adults, 10 adult subjects with established cardiac diseases but maintained right ventricular ejection fraction, and 10 adult subjects with established cardiac diseases and reduced right ventricular ejection fraction).

  • Data Provenance: Retrospective, selected from a comprehensive clinical database of 811 subjects with available echocardiographic data. (No specific country of origin is mentioned, but "clinical database" suggests real-world patient data).

• Database Validation: 811 subjects.

  • Data Provenance: Retrospective 3D echocardiographic clinical database. (No specific country of origin is mentioned).

• cMRI Validation: 3 subjects.

  • Data Provenance: Retrospective Cardiac Magnetic Resonance Imaging (cMRI) data. (No specific country of origin is mentioned).

• Usability Evaluation: 15 participants.

  • Data Provenance: Not specified, but generally, usability testing involves representative users in a simulated environment rather than patient data.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications

• Accuracy Validation: 3 expert cardiologists.

  • Qualifications: "with expertise in 3D echocardiography."

• Database Validation: The comparison was against the predicate device (TomTec-derived values), implying that the predicate acted as a form of ground truth or reference standard for comparison. The document does not describe expert adjudication for these 811 cases beyond the predicate's output.

• cMRI Validation: The comparison was against cMRI, which is often considered a gold standard for cardiac imaging. No human expert count or specific qualifications for establishing ground truth from cMRI are mentioned in the context of this validation, as cMRI itself serves as the reference.

• Usability Evaluation: Not applicable, as this was a usability test, not an accuracy test requiring medical ground truth.

4. Adjudication Method for the Test Set

• Accuracy Validation: The text states, "The cardiologists manually performed segmentation and contouring on the same 3D models that were analyzed by ReVISION Software." This implies an independent assessment by each expert, with the acceptance criteria defining the allowable difference between the device and these manual measurements. There is no explicit mention of an adjudication process (e.g., 2+1, 3+1 consensus) if experts disagreed; it implies each expert's manual measurement was a "ground truth" to be compared against for those specific acceptance criteria. For the visual verification of decomposed ejection fractions, it states "the expert cardiologists' task was to verify visually," again implying individual expert assessment.

• Database Validation: No adjudication method mentioned; comparison was directly between device output and predicate device output.

• cMRI Validation: No adjudication method mentioned; comparison was directly between device output and cMRI measurements.

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 such MRMC comparative effectiveness study involving human readers with and without AI assistance was mentioned or described in the provided text. The studies focused on the standalone performance and accuracy of the device against manual measurements, predicate device measurements, or cMRI.

6. If a Standalone (i.e. algorithm only without human-in-the-loop performance) was done

Yes, extensive standalone performance testing was done.

• Unit/Integration Testing, System Level Testing, Accuracy Validation, Database Validation, and cMRI Validation all describe studies where the Argus Cognitive ReVISION Software's output was directly compared against reference standards (manual expert measurements, predicate device output, or cMRI) without a human-in-the-loop scenario explicitly described to modify or interpret the AI's initial output for the purpose of the performance metrics. The device's "accuracy" was assessed independently.

7. The type of ground truth used

The ground truth used varied depending on the validation test:

• Expert Consensus/Manual Measurements: For "Accuracy Validation," the ground truth for segmental volumes, EFs, and strains was established by 3 expert cardiologists performing manual segmentation and contouring. For decomposed EFs, it was visual verification by expert cardiologists.
• Predicate Device Data: For "Database Validation," the predicate device's calculated parameters (EDV, ESV, SV, EF) served as the reference standard for comparison.
• Outcomes Data: Not explicitly mentioned as a ground truth. The focus is on measurement accuracy against established methods.
• Pathology: Not mentioned.
• cMRI: For "cMRI Validation," Cardiac Magnetic Resonance Imaging (cMRI) measurements were used as a reference for RV volumes and ejection fraction.

8. The Sample Size for the Training Set

The document does not explicitly state the sample size for the training set used to develop the Argus Cognitive ReVISION Software. The description of the device mentions a "rule-based, non-AI/ML algorithm" for initial segmentation, and while the name "Argus Cognitive" implies AI/ML, the performance testing focuses on the resulting calculations rather than the development or training specifics.

9. How the Ground Truth for the Training Set was Established

Since the document predominantly describes the device's segmentation as "rule-based, non-AI/ML algorithm," the concept of a "training set" and "ground truth for the training set" as typically understood in machine learning might not directly apply in the same way for these core components. For rule-based systems, the "ground truth" during development involves expert-defined rules and specifications, which are then verified through unit and integration testing against known geometric data and clinical cases, as described in the "Mathematical Unit Tests" and "Verification Tests" sections. If there are AI/ML components (which the company name suggests), the process of establishing ground truth for those training sets is not detailed in this submission.

Ask a specific question about this device