(103 days)
EndoNaut is an image fusion software solution and computerized navigational system intended to assist X-ray fluoroscopy-guided procedures in the positioning of surgical instruments and endovascular devices.
EndoNaut is indicated for use by Physicians for patients undergoing a fluoroscopy X-ray guided procedure in the chest, abdomen, pelvis, neck and lower limbs, such as aneurysm repair, artery/vein embolization, or peripheral artery disease treatment.
The information provided by the software or system is in no way intended to substitute for, in whole or in part, the surgeon's judgment and analysis of the patient's condition.
It is mandatory to check the real-time anatomy with a suitable imaging technique, such as a contrast-enhanced angiography, before deploying any invasive medical device.
EndoNaut is a computerized navigation system consisting of a software part that carries the medical features and technologies that are controlled and can be installed on a hardware part that enables the medical device to be used in accordance with its intended purpose.
EndoNaut is intended to assist X-ray fluoroscopy-guided procedures in the positioning of surgical instruments and endovascular devices.
EndoNaut Software parts are supported by hardware and software accessories which enable image display and an interaction with the user.
The Software part is interoperable with EndoSize which is a standalone Software designed and developed by Therenva to enable case planning strategy and device (endoprosthesis) selection be-fore endovascular procedure. EndoSize is used by practitioners (in the preparation phase of the operating procedure) or by endoprosthesis manufacturers to visualize vascular structures and/or carry out an extract of the vascular structure from the preoperative CT scan. EndoSize is medical device software which obtained a substantial equivalence determination and FDA clearance through the CDRH premarket notification process (510(K)) (NºK160376).
The provided text is a 510(k) Summary for the medical device EndoNaut. It details the device, its intended use, and a comparison to a predicate device to establish substantial equivalence. However, it does not contain information about a specific study proving the device meets acceptance criteria related to its performance in terms of AI model accuracy, such as sensitivity, specificity, or reader study outcomes.
The document primarily focuses on demonstrating that the new version of EndoNaut (Subject Device) has similar intended use, functionalities, and safety/effectiveness profiles as its previously cleared predicate device (Predicate Device [K212383](https://510k.innolitics.com/search/K212383)), despite some architectural changes (e.g., from standalone software to a server/client model). It emphasizes design verification and validation activities rather than clinical performance studies for AI accuracy.
Here's a breakdown of why the requested information cannot be fully extracted and what can be inferred:
Unable to provide a table of acceptance criteria and reported device performance related to AI model accuracy. The document states:
- "The subject of this premarket submission did not require clinical studies to support equivalence." (Page 17)
- It mentions "Verification and validation activities have demonstrated that the EndoNaut (Server) software variant performs equally as the EndoNaut (Standalone) predicate software by providing reliable results, without functional regression and moreover, offers robust safety/security mechanisms." (Page 18)
- It refers to "Features which call machine-learning algorithms: Registration 3D/2D Motion detection Contrast injection detection" (Page 6). It states that "[t]he algorithms have not been changed. Only the way they are implemented is different... These implementation changes do not change the purpose of the algorithms (they do what they did before)." (Page 6)
This implies that the "performance" validated was primarily related to the continued functionality and safety of the existing algorithms within the new architecture, rather than a re-evaluation of their diagnostic accuracy or impact on human performance in a clinical setting. Thus, there are no specific accuracy metrics (e.g., sensitivity, specificity, AUC) or reader study results reported for the device in this document.
Based on the provided text, the following information is either explicitly stated, can be inferred, or is explicitly absent:
-
A table of acceptance criteria and the reported device performance:
- Acceptance Criteria for AI Performance (e.g., sensitivity, specificity): Not provided in the document. The submission focuses on substantial equivalence to a predicate device and verification of existing algorithms within a new architecture, rather than establishing new performance benchmarks for AI accuracy.
- Reported Device Performance: No quantitative performance metrics (e.g., sensitivity, specificity, accuracy, AUC) related to the AI algorithms' diagnostic capabilities are reported. The performance discussion centers on "functional regression" checks and demonstrating "reliable results" and "robust safety/security mechanisms" for the software change (Page 18).
-
Sample size used for the test set and the data provenance: Not specified in the context of AI performance evaluation. The document mentions "Simulated use testing (Validation)" (Page 18) but does not detail the dataset used, its size, or provenance.
-
Number of experts used to establish the ground truth for the test set and the qualifications of those experts: Not specified. As no clinical studies or specific accuracy evaluations are detailed, the ground truth establishment for such performance metrics is not discussed.
-
Adjudication method (e.g., 2+1, 3+1, none) for the test set: Not specified.
-
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: Not done/reported. The document explicitly states: "The subject of this premarket submission did not require clinical studies to support equivalence." (Page 17).
-
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done: Not explicitly detailed with performance metrics. The document confirms the presence of machine learning algorithms for "Registration 3D/2D, Motion detection, Contrast injection detection" and states that "The algorithms have not been changed. Only the way they are implemented is different... These implementation changes do not change the purpose of the algorithms (they do what they did before)." (Page 6). This suggests that the algorithms themselves were part of the predicate device and their function was maintained, but a formal standalone performance study for this specific submission is not presented. The verification activities might have included internal functional testing of these algorithms.
-
The type of ground truth used (expert consensus, pathology, outcomes data, etc.): Not specified regarding the AI features. For the general functionality, the "ground truth" would be the expected behavior and output based on design specifications and the predicate device's performance.
-
The sample size for the training set: Not specified. The document states that the machine learning algorithms "have not been changed" from the predicate device (Page 6), implying that any training would have occurred for the predicate device, but details are not provided here.
-
How the ground truth for the training set was established: Not specified.
Summary based on the document:
This 510(k) submission for EndoNaut (K222070) is primarily concerned with demonstrating substantial equivalence to a previously cleared predicate device (EndoNaut K212383) following changes in software architecture and minor hardware updates. It relies on:
- Design verification and validation (V&V) activities: These include risk assessments, usability reviews, requirement reviews, design reviews, clinical evaluation report reviews, testing on unit level, integration testing, interoperability testing, performance testing, safety testing, and simulated use testing (Page 18).
- No new clinical studies: The submission explicitly states that clinical studies were not required to support equivalence (Page 17).
- Consistency of AI algorithms: The document clarifies that the underlying machine learning algorithms for features like 3D/2D registration, motion detection, and contrast injection detection have not been changed but their implementation differs due to the new software architecture. The validation activities focused on ensuring these algorithms perform "equally as the EndoNaut (Standalone) predicate software" (Page 18) without "functional regression" and maintaining safety and security.
Therefore, the document does not provide the detailed acceptance criteria or performance study results typically associated with AI model accuracy (e.g., sensitivity, specificity, reader studies) for a new AI function or a significant modification that would necessitate re-evaluation of diagnostic performance. Instead, it attests to the maintained functionality and safety of existing features within an updated system.
§ 892.1650 Image-intensified fluoroscopic x-ray system.
(a)
Identification. An image-intensified fluoroscopic x-ray system is a device intended to visualize anatomical structures by converting a pattern of x-radiation into a visible image through electronic amplification. This generic type of device may include signal analysis and display equipment, patient and equipment supports, component parts, and accessories.(b)
Classification. Class II (special controls). An anthrogram tray or radiology dental tray intended for use with an image-intensified fluoroscopic x-ray system only is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 892.9. In addition, when intended as an accessory to the device described in paragraph (a) of this section, the fluoroscopic compression device is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 892.9.