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

Biolmagene PATHIAM is intended for use as an accessory to the Dako HercepTest® to aid a pathologist in semi-quantitative measurement of HER2/neu (c-erbB-2) in Formalin-fixed, paraffin-embedded breast cancer tissue. When used with the Dako HercepTest it is indicated as an aid in the assessment of breast cancer patients for whom Herceptin® (Trastuzumab) treatment is being considered.

The imaging software is intended to detect and classify cells of clinical interest by analyzing digitized images of microscope slides based on recognition of cellular objects of particular color, size and shape. The software can be used with a computer and image digitizer with features specified in the labeling.

Device Description

PATHIAM software is a standalone software application that will work on a system with the following features required but not provided:

Computer

Processor: 2.4 GHz, Pentium IV equivalent
Memory: 512 MB RAM
Operating System: Windows 2000 or later
Hard Drive: minimum 100MB for software installation, 20GB for image storage
LAN connectivity, minimum 100 MBPS (recommended), support for USB interface, support for HTTP, TCP/IP protocols (using the Operating system)
High Speed Graphic Accelerator Card (1024 X 768)
17" High resolution display monitor
24 bit color depth
Font Setting: Small font (DPI setting: 96 DPI)

Digitizing Equipment: Camera

Resolution: at least 2048 x 1536 pixels
Frame rate: 20 fps@1200 x 768 resolution (6 fps @ 2048 x 1536 resolution)
Sampling Frequency of 6.26 square/um
Compression format: JPEG 2000, BMP, TIFF, JPEG
Color: 24-bit (R. G. B)
Connection to computer

Digitizing Equipment: Digital Side Scanner

Input Format: 25X75mm microscope slides
Resolution: 54,000 pixel/inch with 20X objective
Method: Line-scanning
File Format: TIFF/JPEG2000; compliant with TIFF 6.0 standard.
Color: 24-bit (R.G. B)
Connectivity: 100/1000 MBPS Ethernet
Compression format: JPEG 2000, BMP, TIFF, JPEG

The software allows both archiving of the digital image, and semi quantitative analysis of extent and intensity of stained tissue, providing the pathologist with an aid to interpretation of level of expression of Her2/neu in breast cancer tissue. The pathologist is presented with a digital image of the tissue section and a suggested staining score (0 to 3). The pathologist then makes an assessment of the digital image and reports his/her score.

AI/ML Overview

Here's a breakdown of the acceptance criteria and study details for the BioImagene PATHIAM Image Analysis Software for Her2/neu, based on the provided 510(k) summary:

Acceptance Criteria and Reported Device Performance

The acceptance criteria for the PATHIAM software are implicitly defined by comparison to the established predicate device (ACIS Her2/neu software component) and through performance studies demonstrating inter-laboratory and inter-reader agreement. While explicit numerical acceptance criteria for overall agreement are not stated as "target thresholds," the demonstrated high levels of agreement strongly support the device's performance.

Criterion	Acceptance Criteria (Implied)	Reported Device Performance
Between-laboratory reproducibility for PATHIAM raw scores (Automated analysis only)	High overall agreement across different labs/imaging systems.	94% - 95% Overall Agreement (95% CI: 89-98%) among three different labs. (Table 1)
Between-laboratory reproducibility for Pathologist Assisted by PATHIAM Scores	High overall agreement across different labs with pathologist assistance.	81% - 96% Overall Agreement (95% CI: 74-98%) among three different labs. (Table 2)
Comparison of PATHIAM-Assisted vs. Manual Scores (Intra-lab agreement)	High agreement between a pathologist's score with and without PATHIAM assistance from the same lab.	81% - 84% Overall Agreement (95% CI: 75-89%) for each of the three labs. (Table 4)
Comparison of PATHIAM raw scores vs. Manual Scores (Intra-lab agreement)	High agreement between the raw algorithm score and a pathologist's manual score from the same lab.	78% - 83% Overall Agreement (95% CI: 71-88%) for each of the three labs. (Table 5)
Substantial Equivalence to Predicate Device	Similar intended use, indications for use, specimen type, image analysis system, method of cell detection, hardware components, and assay used.	The device is compared favorably to the predicate (ACIS Her2/neu software component) across these attributes, indicating substantial equivalence. (Table 6)

Study Details

Sample sizes used for the test set and the data provenance:
- Test Set Sample Size: 176 stained breast cancer tissue specimens.
- Data Provenance: The study was conducted in the US at three different sites. There is no explicit mention of the data being retrospective or prospective, but the description of "analyzed images of the same set of 176 stained breast cancer tissue specimens" suggests a retrospective analysis of existing samples.
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- Number of Experts: Three different pathologists.
- Qualifications: They are referred to as "trained pathologists," but no specific experience levels (e.g., years of experience, board certifications) are provided.
Adjudication method (e.g. 2+1, 3+1, none) for the test set:
- The study involved multiple pathologists providing scores, and comparisons were made between their scores. However, there is no explicit mention of an adjudication method to establish a single, definitive ground truth score for each case through consensus or a tie-breaker. Instead, the study focuses on agreement between the individual scores of the device, manual readers, and device-assisted readers across different labs.
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:
- Yes, implicitly, an MRMC comparative effectiveness study was performed in the sense that multiple readers (pathologists) evaluated multiple cases with and without the PATHIAM assistance, and across different laboratory setups.
- Effect Size of Improvement:
  - Manual Scores (Inter-lab agreement): 76% - 97% overall agreement (Table 3).
  - Pathologist Assisted by PATHIAM Scores (Inter-lab agreement): 81% - 96% overall agreement (Table 2).
  - The summary states: "Consistency is improved when the PATHIAM score assists the pathologist in their interpretation (Table 2)."
  - While not quantified as a single "effect size" number, comparing Table 3 (Manual Scores) to Table 2 (PATHIAM-Assisted Scores) suggests an improvement in agreement. For instance, Lab 1 vs Lab 3 manual agreement was 76%, while with PATHIAM assistance, it was 81%. Lab 2 vs Lab 3 manual agreement was 78%, and with PATHIAM assistance, it was also 81%. This indicates a trend towards improved inter-reader consistency with AI assistance.
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:
- Yes, a standalone performance assessment was conducted. This is evidenced by the "Between-Lab Agreement for Raw PATHIAM- Scores" (Table 1), which shows the consistency of the algorithm's output across different laboratories without direct pathologist modification of the software's initial score. The study also compared "PATHIAM raw scores and Manual Scores" (Table 5), which is a comparison of the algorithm's output against human interpretation.
The type of ground truth used (expert consensus, pathology, outcomes data, etc.):
- The primary "ground truth" or reference for comparison in this study is the pathologist's interpretation, both manual (microscopic assessment) and PATHIAM-assisted. The study focuses on agreement between these interpretations rather than against an external, independent gold standard like pathology results or clinical outcomes. The device is an aid to a pathologist, implying its performance is assessed by how well it aligns with or improves human expert interpretation.
The sample size for the training set:
- The document does not provide information on the sample size used for the training set. The performance data presented relates exclusively to the test set used for validation.
How the ground truth for the training set was established:
- Since the training set size and details are not provided, it is also not stated how the ground truth for any potential training set was established.

Summary

Regulation Number and Section

§ 864.1860 Immunohistochemistry reagents and kits.

(a)
Identification. Immunohistochemistry test systems (IHC's) are in vitro diagnostic devices consisting of polyclonal or monoclonal antibodies labeled with directions for use and performance claims, which may be packaged with ancillary reagents in kits. Their intended use is to identify, by immunological techniques, antigens in tissues or cytologic specimens. Similar devices intended for use with flow cytometry devices are not considered IHC's.(b)
Classification of immunohistochemistry devices. (1) Class I (general controls). Except as described in paragraphs (b)(2) and (b)(3) of this section, these devices are exempt from the premarket notification requirements in part 807, subpart E of this chapter. This exemption applies to IHC's that provide the pathologist with adjunctive diagnostic information that may be incorporated into the pathologist's report, but that is not ordinarily reported to the clinician as an independent finding. These IHC's are used after the primary diagnosis of tumor (neoplasm) has been made by conventional histopathology using nonimmunologic histochemical stains, such as hematoxylin and eosin. Examples of class I IHC's are differentiation markers that are used as adjunctive tests to subclassify tumors, such as keratin.(2) Class II (special control, guidance document: “FDA Guidance for Submission of Immunohistochemistry Applications to the FDA,” Center for Devices and Radiologic Health, 1998). These IHC's are intended for the detection and/or measurement of certain target analytes in order to provide prognostic or predictive data that are not directly confirmed by routine histopathologic internal and external control specimens. These IHC's provide the pathologist with information that is ordinarily reported as independent diagnostic information to the ordering clinician, and the claims associated with these data are widely accepted and supported by valid scientific evidence. Examples of class II IHC's are those intended for semiquantitative measurement of an analyte, such as hormone receptors in breast cancer.
(3) Class III (premarket approval). IHC's intended for any use not described in paragraphs (b)(1) or (b)(2) of this section.
(c)
Date of PMA or notice of completion of a PDP is required. As of May 28, 1976, an approval under section 515 of the Federal Food, Drug, and Cosmetic Act is required for any device described in paragraph (b)(3) of this section before this device may be commercially distributed. See § 864.3.