(124 days)
The VX1 assists operators in the real-time manual annotation of 3D anatomical and electrical maps of human atria for the presence of multipolar intra-cardiac atrial electrograms exhibiting spatiotemporal dispersion during atrial fibrillation or atrial tachycardia.
The clinical significance of utilizing the VX1 software to help identify areas with intra-cardiac atrial electrograms exhibiting spatiotemporal dispersion for catheter ablation of atrial arrhythmias, such as atrial fibrillation, has not been established by clinical investigations.
The VX1 "" is a machine and deep learning based-algorithm designed to assist operators in the real-time manual annotation of 3D anatomical and electrical maps of the human atria for the presence of electrograms exhibiting spatiotemporal dispersion during atrial fibrillation or atrial tachycardia", i.e., dispersed intra-cardiac atrial electrograms (DEs). The device works with all existing 510(k) cleared catheters that meet specific dimension requirements and with two data acquisition systems: the LabSystem Pro Acquisition System (Boston Scientific) (K141185) and the MacLab CardioLab Acquisition System (General Electric) (K130626). A connection cable is used to connect the data acquisition system with an Advantech PCI-1713U analog-to-digital converter, which transmits the acquired information to a nearby computer that hosts the VX1 software. The computer and its attached display are located outside the sterile operating room area. The VX1 software analyzes the patient's electrograms to cue operators in real-time to intra-cardiac electrograms of interest for atrial regions harboring DEs as well as a cycle length estimation from electrograms recorded with the mapping and the coronary sinus catheters. The results of the analysis are graphically presented on the attached computer display.
-
Table of Acceptance Criteria and Reported Device Performance:
The document does not explicitly present a table of acceptance criteria with corresponding performance values in the way typically seen in a formal statistical analysis plan. However, the outcomes of the "Reader Study" and the "Limited Time Annotation Testing" serve as a form of acceptance criteria, where the device's agreement with expert consensus is evaluated.
Performance Metric Acceptance Criteria (Implied) Reported Device Performance Inter-operator agreement Strong agreement among expert readers Confirmed. Experts strongly agreed with one another. VX1 correlation with expert annotation (Reader Study) Acceptable correlation with physician annotation VX1 adjudication acceptably correlates with physician annotation. Period estimation performance Outperform or be comparable to alternative algorithms VX1 period estimation algorithm outperformed both FFT-based and AC-based algorithms for noisy/non-periodic electrograms. VX1 algorithm accuracy (k-fold cross-validation) High accuracy Over 89% accuracy. VX1 performance vs. unassisted electrophysiologists Superior agreement to annotating experts than unassisted humans VX1 algorithm demonstrated superior performance (agreement to annotating experts) to cardiac electrophysiologists. Clinical risk and procedure time No additional risk or procedure time Not associated with additional risk or procedure time. Comparable performance to predicate's CFAE module Comparable performance VX1 device demonstrated comparable performance in locating electrical signal transmission pattern disruptions as the predicate's CFAE module. -
Sample sizes used for the test set and data provenance:
- Reader Study: 14,370 electrograms.
- Period Estimation Algorithm Testing: 2,550 electrograms.
- Limited Time Annotation Testing: 1,020 electrograms.
- Clinical Study: 300 patients across 8 centers.
- Data Provenance: The document states "An OUS clinical study was performed..." indicating that at least some clinical data is from outside the US. The "training database of 275,020 1.5 second annotated electrograms" drew on "anonymized information from a very large database". The nature (retrospective/prospective) is not explicitly stated for all datasets, but the clinical study was likely prospective to evaluate usability and efficacy.
-
Number of experts used to establish the ground truth for the test set and qualifications of those experts:
- Reader Study and Limited Time Annotation Testing: 3 independent experts were used to adjudicate the 14,370 electrograms for the Reader Study. For the Limited Time Annotation Testing, 2 annotating experts were used (who reached a consensus). Their specific qualifications beyond "independent experts" and "annotating experts" are not detailed (e.g., years of experience or board certification).
- "Unassisted" Human Readers in Limited Time Annotation Testing: 28 cardiac electrophysiologists who underwent a "full-day seminar on spatiotemporal dispersion."
-
Adjudication method for the test set:
- Reader Study: "adjudicated by three independent experts". It doesn't specify if this was a 2+1, 3+1, or simple majority.
- Limited Time Annotation Testing: "two annotating experts who were allowed unlimited time to make a classification determination and reach a consensus". This suggests a consensus-based adjudication between two experts.
-
If a multi-reader multi-case (MRMC) comparative effectiveness study was done, and the effect size:
- A comparative study was done during the "Limited Time Annotation Testing" comparing the VX1 algorithm's performance to that of 28 cardiac electrophysiologists. This fits the description of a multi-reader (28 electrophysiologists) multi-case (1,020 electrograms) study.
- Effect Size: The document states the "VX1 algorithm demonstrated superior performance (agreement to the annotating experts) to the cardiac electrophysiologists in identifying dispersed and non-dispersed intra-cardiac atrial electrograms." However, a specific effect size (e.g., AUC difference, sensitivity/specificity improvement, or statistical significance value) is not provided.
-
If a standalone performance (algorithm only without human-in-the-loop performance) was done:
- Yes, the "Reader Study" directly compared the VX1's classification to expert annotation, implying a standalone performance evaluation.
- The "Period estimation algorithm testing" also evaluated the algorithm's performance independently.
- The k-fold cross-validation is a standalone algorithm performance test.
-
The type of ground truth used:
- Expert Consensus: The primary ground truth for the Reader Study and Limited Time Annotation Testing was established by "independent experts" or "annotating experts" reaching a consensus on the presence or absence of dispersed intra-cardiac atrial electrograms.
-
The sample size for the training set:
- The k-fold cross-validation was performed on a "training database of 275,020 1.5 second annotated electrograms."
-
How the ground truth for the training set was established:
- The training database drew on "anonymized information from a very large database... of 1.5 second snippets of multipolar intra-cardiac atrial electrograms located on a dedicated data server". The document does not explicitly state how these 275,020 electrograms were annotated (e.g., by experts, or through a different automated process). However, the context of comparing the algorithm to expert annotations in the validation suggests that expert annotation would be the most robust method for establishing ground truth for training as well. Without explicit mention, it's inferred to be expert-annotated.
§ 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).