FDA 510(k) Search - By Innolitics (SaMD and AI Experts)

The Fluoptics Fluobeam® Imaging system is intended to provide real-time near infrared (NIR) fluorescence imaging of tissue during surgical procedures. The Fluoptics Fluobeam® Imaging system is indicated for use in capturing and viewing fluorescent images for the visual assessment of blood flow in adults as an adjunctive method for the evaluation of tissue perfusion, perfused organs, and related tissue-transfer circulation in tissue and free flaps used in plastic, micro- and reconstructive and organ transplant surgeries.

The Fluoptics Fluobeam® Imaging system can also be used to assist in the imaging of parathyroid glands and can be used as an adjunctive method to assist in the location of parathyroid glands due to the auto-fluorescence of this tissue.

Use of the Fluobeam® device is intended to assist. not replace, experienced visual assessment, and biopsy with conventional histopathological confirmation per standard of care. The system is not to be used to confirm the absence of parathyroid tissue or glands and is only to be used to assist in location of visually identified gland/tissues.

Device Description

The Fluobeam 800 Clinic Imaging Device Used With Fluocase 800 Control System is an autofluorescence imaging system that is capable of visualizing autofluorescent signals from the parathyroid glands. The device is a non-contacting imaging system that excites fluorescent molecules with non-ionizing near-infrared light at 750 nm and collects emissions from 800 nm to (6) (4) mm. The collected emissions are subsequently displayed as an image on a panel PC screen.

The Fluobeam device is composed of the following components:

1. The optical head (FluoBeam 800 Clinic® Device)
- a. Contains 750 nm laser (for fluorescence excitation), NIR LEDs (b) (4) and white LEDs (normal illumination λ

AI/ML Overview

The Fluoptics Fluobeam 800 Clinic® Imaging Device with Fluocase 800™ Control System is an autofluorescence detection device for general surgery and dermatological use. It is intended to provide real-time near-infrared (NIR) fluorescence imaging of tissue during surgical procedures for the visual assessment of blood flow and as an adjunctive method to assist in the localization of parathyroid glands due to their autofluorescence. It is explicitly stated that the device is to assist, not replace, experienced visual assessment and biopsy with conventional histopathological confirmation, and is not to be used to provide a diagnosis or confirm the absence of parathyroid tissue.

Here's an analysis of its acceptance criteria and the supporting studies:

1. Table of Acceptance Criteria and Reported Device Performance

The provided text does not explicitly define a set of quantitative "acceptance criteria" for the device's performance in terms of sensitivity, specificity, or other metrics (e.g., a specific minimum sensitivity for parathyroid detection). Instead, the acceptance is based on demonstrating the device's clinical applicability as an adjunctive tool and the mitigation of identified risks. The special controls listed define requirements for performance testing, but not quantitative thresholds.

However, based on the Clinical Conclusions and the Benefit-Risk Determination, the implicit acceptance criteria are that the device can:

Consistently demonstrate autofluorescence of parathyroid glands with an intensity typically greater than surrounding tissues.
Improve parathyroid gland localization during surgery.
Result in reduced adverse clinical outcomes such as postoperative hypocalcemia, inadvertent resection, and autotransplantation when used as an adjunct.
Demonstrate safety, with identified risks mitigated through performance testing, software verification, and labeling.
Not lead to significant increases in operative time.

Here's how the reported device performance addresses these implicit criteria:

Acceptance Criterion (Implicit)	Reported Device Performance (from Clinical Conclusions)	Supporting Study
Consistent autofluorescence of parathyroid glands with higher intensity than surrounding tissues.	Parathyroid glands consistently autofluoresce with average intensity typically greater than nearby and surrounding tissues. (Mean intensity for parathyroid glands was 40.6 (±26.5) vs. thyroid 31.8 (±22.3) and background 16.6 (±15.4) in Study 1; Mean intensity for parathyroid glands was 47.6 (±26.9) vs. thyroid 22.2 and background 9.1 in Study 2). 98% of ultimately identified parathyroid glands autofluoresced in Study 4.	Study 1, Study 2, Study 4
Improve parathyroid gland localization during surgery.	Autofluorescence appears to allow detection of parathyroid glands earlier in the surgical procedure. The number of parathyroid glands visualized with the device was significantly higher than with direct direct light in Study 2 (mean 3.7 vs 2.5). In Study 3, parathyroid identification rates were higher in the NIR+ group compared to NIR- group (76.3% vs. 65.7% of theoretically present parathyroids). In Study 4, 46% of located glands were not identified on initial visual inspection, and in 77% of patients, at least one gland was detected by autofluorescence before direct inspection.	Study 2, Study 3, Study 4
Reduction in adverse clinical outcomes (postoperative hypocalcemia, inadvertent resection, autotransplantation) when used as an adjunct.	Study 3 provides reasonable affirmation that earlier detection can result in reduced postoperative transient hypocalcemia (NIR+ 5.3% vs NIR- 20.9%, Control 1 16.1%, Control 2 19.5%), inadvertent resection (NIR+ 1.1% vs NIR- 7.2%, Control 1 8%, Control 2 6.9%), and autotransplantation (NIR+ 2.1% vs NIR- 15.0%, Control 1 16.7%, Control 2 16.1%).	Study 3
Demonstrates safety (mitigation of electrical, mechanical, thermal, light/laser, infection, adverse tissue reaction, false identification risks).	Passes various safety standards including IEC 60601-1, IEC 60601-1-2, IEC 60601-1-6, IEC 60825-1, EN 62471. Software verification and validation performed (moderate level of concern). No direct/indirect patient contacting components for biocompatibility. Shelf life and sterility addressed for sterile sheath. Labeling includes risk mitigation warnings for false identification. No device-related AEs, SAEs, or UADEs reported in clinical studies.	Non-clinical/Bench Studies, Software, Labeling, Risks to Health
No significant increase in operative time.	Study 3 reported no significant difference in operative time for the same surgeon with and without the device (NIR+ compared to NIR-).	Study 3
Acceptable occurrence of false positives/negatives when used as an adjunct, with appropriate labeling. (The device is not diagnostic and not for confirmation or concluding absence of parathyroid tissue, so absolute accuracy metrics are not the primary goal).	False negatives are not frequent (98% reported sensitivity in Study 4 regarding identified parathyroid glands). False positives represent a more frequent occurrence (13 glands in 28 patients were false positives in Study 1; colloid nodules in Study 5). This moderate uncertainty of false diagnoses risk is mitigated by adjunctive use and labeling indicating the device is not for confirmation or diagnosis.	Study 1, Study 4, Study 5, Labeling, Benefit-Risk

2. Sample size used for the test set and the data provenance

Study 1 (Falco et al 2016):

Sample Size: 28 patients.
Data Provenance: Prospective, single institution (unspecified country, but given authors are from US/Argentina in other papers, likely US or Europe), between June 2015 and August 2015.

Study 2 (Falco et al 2017):

Sample Size: 74 patients.
Data Provenance: Prospective, single institution (unspecified country, likely same as Study 1), between October 2015 and February 2016.

Study 3 (Benmiloud et al):

Sample Size: 513 total patients (NIR+ group: 93, NIR- group: 153, Control 1: 180, Control 2: 87).
Data Provenance: "Before and After Controlled Study", single institution (unspecified country, authors typically based in France). Data from January 2015-January 2016 (Period 1) and February 2016-September 2016 (Period 2).

Study 4 (Kahramangil et al):

Sample Size: 210 prospectively-enrolled patients.
Data Provenance: Retrospective review of prospectively-enrolled patients from three centers (one being the same hospital as Study 3). Unspecified years for data acquisition, unspecified country (authors typically based in US/Argentina/Turkey/France).

Study 5 (De Leeuw et al):

Sample Size: 35 patients (28 specimens included for analysis).
Data Provenance: Prospective, single-center investigation (unspecified country, authors from France). Data between December 2014 and March 2015.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

Study 1: Histologic confirmation for adenomas and normal gland biopsies for patients with primary hyperparathyroidism. Other surgical procedures did not have histology confirmation. The specific number and qualifications of pathologists are not mentioned.
Study 2: Parathyroid adenomas were resected for histology. Normal glands were not resected for histology. The specific number and qualifications of pathologists are not mentioned.
Study 3: "Parathyroids were only recorded as observed if the surgeon had 'no doubts' that the tissue was parathyroid." No tissue biopsies were performed for confirmation in the groups. The surgeons were "Surgeon 1 (five years of experience)" and "Surgeon 2 (twenty-five years of experience)."
Study 4: Parathyroids were confirmed with either frozen section histology, or if they met three visual criteria (yellow brown color, discrete shape, distinct vasculature). The specific number and qualifications of pathologists or surgeons are not mentioned beyond the general description.
Study 5: A "blinded pathologist" identified tissue using conventional histology. The qualifications or specific number of pathologists are not given.

4. Adjudication method for the test set

Study 1: No explicit adjudication method mentioned. Histologic confirmation used where possible.
Study 2: No explicit adjudication method mentioned. Histologic confirmation used for adenomas.
Study 3: The surgeon's "no doubts" visual assessment was the primary "ground truth" for identified parathyroids, coupled with clinical outcomes. No independent adjudication mentioned.
Study 4: Confirmation by either frozen section histology or by meeting three visual criteria. No separate adjudication process beyond these methods.
Study 5: Histology results from a blinded pathologist were compared to the "scientist's" determination using Fluoptics. The "scientist" was a blinded investigator (not a clinician, unfamiliar with anatomy). This method itself acts as a form of adjudication against a histological ground truth.

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 explicit traditional Multi-Reader Multi-Case (MRMC) comparative effectiveness study evaluating human readers' improvement with vs without AI assistance was described in the provided text. The studies focused on surgeon performance with and without the device, but not in a strict MRMC design with defined reader cohorts analyzing cases.

Study 2 did compare visualizations using the device (NIRL) versus direct inspection (WL), showing a mean difference (NIRL-WL) of 1.2 (0.8) parathyroid glands per patient, with 86.5% of patients having four glands visualized with the device vs. 12.2% with white light. This suggests an improvement in glandular visualization.
Study 3 compared clinical outcomes between groups where a surgeon did or did not use the device. Autotransplantation rates were significantly reduced in the NIR+ group compared to all other groups (e.g., NIR+: 2.1% vs. NIR-: 15.0%), and inadvertent parathyroid resection also occurred less frequently. This implies an improvement in surgical performance (a human "reader" or surgeon using the device). The "effect size" can be inferred from the differences in clinical outcomes and identification rates (e.g., reduction in hypocalcemia from 20.9% to 5.3% for transient hypocalcemia in one comparison).

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

No standalone performance of the algorithm (device processing without human interpretation) was explicitly assessed or reported. The device is consistently described as an adjunctive tool to assist human surgeons. Clinical studies always involve a surgeon using the device in real-time.

7. The type of ground truth used

The ground truth varied across studies:

Study 1 & 2: Histology for resected parathyroid adenomas and normal gland biopsies (where performed). For many other tissues, visual assessment by the surgeon.
Study 3: Surgeon's visual assessment ("no doubts") that tissue was parathyroid, complemented by clinical outcomes (hypocalcemia, autotransplantation, inadvertent resection). No histology.
Study 4: Frozen section histology OR meeting three visual criteria (yellow-brown color, ovoid shape, distinct vasculature) by the surgeon.
Study 5: Conventional histology (H&E Saffron staining) by a blinded pathologist.

8. The sample size for the training set

The provided documents describe clinical studies used for evaluating the device's performance and supporting its regulatory acceptance, not for training a specific AI algorithm. The device, the Fluobeam 800 Clinic, is an imaging system that captures autofluorescent signals and displays them. While it has software with "several modes (standard, advanced, perfusion, low signals, and time lapse) for visualizing fluorescence and autofluorescence images" and allows for adjustment of imaging parameters, it is not described as an AI/ML algorithm that requires a "training set" in the conventional sense (i.e., for learning to identify features or make decisions). It functions more as an imaging modality. Therefore, a "training set" for an AI algorithm is not applicable or provided in the context of this device description.

9. How the ground truth for the training set was established

As noted above, the device is an imaging system, not an AI/ML algorithm requiring a training set. Therefore, this question is not applicable.

Summary

Regulation Number and Section

§ 878.4550 Autofluorescence detection device for general surgery and dermatological use.

(a)
Identification. An autofluorescence detection device for general surgery and dermatological use is an adjunct tool that uses autofluorescence to detect tissues or structures. This device is not intended to provide a diagnosis.(b)
Classification. Class II (special controls). The special controls for this device are:(1) In vivo testing under anticipated conditions of use must characterize the ability of the device to detect autofluorescent signals from tissues or structures consistent with the indications for use.
(2) The patient-contacting components of the device must be demonstrated to be biocompatible.
(3) Performance testing must demonstrate the electromagnetic compatibility and electrical, mechanical, and thermal safety of the device.
(4) Software verification, validation, and hazard analysis must be performed.
(5) Performance testing must demonstrate the sterility of patient-contacting components of the device.
(6) Performance testing must support the shelf life of device components provided sterile by demonstrating continued sterility and package integrity over the labeled shelf life.
(7) Performance testing must demonstrate laser and light safety for eye, tissue, and skin.
(8) Labeling must include the following:
(i) Instructions for use;
(ii) The detection performance characteristics of the device when used as intended; and
(iii) A shelf life for any sterile components.