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

ADAS 3D is indicated for use in clinical settings to support the visualization and analysis of MR and CT images of the heart for use on individual patients with cardiovascular disease.

ADAS 3D is indicated for patients with myocardial scar produced by ischemic or non-ischemic heart disease. ADAS 3D processes MR and CT images. The quality and the resolution of the medical images determines the accuracy of the data produced by ADAS 3D.

ADAS 3D is indicated to be used only by qualified medical professionals (cardiologists, electrophysiologists, radiologists or trained technicians) for the calculation, quantification and visualization of cardiac images and intended to be used for pre-planning and during electrophysiology procedures. The data produced by ADAS 3D must not be used as an irrefutable basis or a source of medical advice for clinical diagnosis or patient treatment. The data produced by ADAS 3D is intended to be used to support qualified medical professionals for clinical decision making.

The clinical significance of using ADAS 3D to identify arrhythmia substrates for the treatment of cardiac arrhythmias (e.g., ventricular tachycardia) or risk stratification has not been established.

Device Description

ADAS 3D is a stand-alone software tool intended to be used for post-processing cardiovascular enhanced Magnetic Resonance (MR) images and Computed Tomography Angiography (CTA) images that are formatted in the Digital Imaging and Communication in Medicine (DICOM) standard. ADAS 3D software aids in the non-invasive calculation, quantification of cardiac imaging data to support a comprehensive diagnostic decision-making process for understanding cardiovascular disease.

ADAS 3D exports information to multiple industry standard file formats suitable for documentation and information sharing purposes. The 3D data is exported into industry standard file formats supported by catheter navigation systems.

ADAS 3D analyses the enhancement of myocardial fibrosis from DICOM MR images to support:

Visualization of the distribution of the enhancement in a three-dimensional (3D) chamber of the heart.
Quantification of the total volume of the enhancement within the Left Ventricle (LV) and the visualization of the enhancement area in multiple layers through the cardiac structure.
. Calculation, quantification and visualization of corridors of intermediate, signal intensity enhancement in the LV.
Quantification and visualization of the total area and distribution of the enhancement within the left Atrium (LA).

Additionally, ADAS 3D imports DICOM CTA images to support:

Quantification of LV wall thickness.
Identification and Visualization of other 3D anatomical structures.
Quantification and visualization of LA wall thickness.
Quantification and visualization of distances from the LA epicardium to other 3D anatomical structures.

Additionally, ADAS 3D imports DICOM Magnetic Resonance Angiography (MRA) images to support:

Identification and Visualization of other 3D anatomical structures.
Additionally, ADAS 3D uses the following machine-learning-based features:
Standard Initialization of the LV, LA, and Aorta from CTA
Standard Initialization of the Coronary Arteries from CTA
Standard Initialization of the LA from CTA
Standard Initialization of the LV from 2D LGE-MRI and Automatic Slice Alignment ●
Standard Initialization of the LV from 3D LGE-MRI
Standard Initialization of the LA from 3D LGE-MRI

AI/ML Overview

Here's a breakdown of the acceptance criteria and study details for the Adas3D Medical S.L. ADAS 3D device, based on the provided text:

Acceptance Criteria and Reported Device Performance

Machine Learning feature	Target structure	Metric	Mean Reported Performance	Threshold	Meets Acceptance Criteria
Standard Initialization of the Left Chambers and Aorta from CTA	LV	DC	0.93	0.84	yes
	LV	MSD	1.29	2.23	yes
	LA	DC	0.94	0.84	yes
	LA	MSD	1.06	2.23	yes
	AO	DC	0.94	0.84	yes
	AO	MSD	0.93	2.23	yes
	LAA	DC	0.84	0.76	yes
	LAA	MSD	1.01	2.23	yes
Standard Initialization of the Coronary Arteries from CTA	LCA	DC	0.82	0.78	yes
	LCA	HD	7.71	10.86	yes
	RCA	DC	0.82	0.78	yes
	RCA	HD	6.61	10.86	yes
Standard Initialization of the LA from CTA	LA ENDO	MSD	0.37	0.32	no
	LA EPI	MSD	0.56	0.76	yes
	LA	CA	56.09	43.90	yes
	LA	CD1	38.32	49.00	yes
	LA	CD2	5.58	12.10	yes
Automatic Slice Alignment for LV from 2D LGE-MRI	LV	MDS	2.39	6.23	yes
Standard Initialization of the LV from 2D LGE-MRI	LV ENDO	DC	0.90	0.85	yes
	LV ENDO	APD	2.01	2.10	no
	LV ENDO	HD	9.72	13.25	yes
	LV EPI	DC	0.93	0.89	yes
	LV EPI	APD	2.06	1.93	no
	LV EPI	HD	9.81	13.25	yes
Standard Initialization of the LV from 3D LGE-MRI	LV ENDO	DC	0.88	0.79	yes
	LV ENDO	HD	2.40	27.32	yes
	LV EPI	DC	0.91	0.78	yes
	LV EPI	HD	9.57	27.32	yes
Standard Initialization of the LA from 3D LGE-MRI	LA	DC	0.90	0.86	yes
	LA	MSD	1.62	1.39	no
	LA	HD	12.36	16.50	yes

Note: The text explicitly states that four tests did not meet the non-inferiority criteria, but these discrepancies were considered sub-pixel and acceptable.

Study Details

Sample size used for the test set and the data provenance:
- Standard Initialization of the Left Chambers and Aorta from CTA: 100 cases (US 62%, OUS 38%)
- Standard Initialization of the Coronaries from CTA: 100 cases (US 64%, OUS 36%)
- Standard Initialization the LA from CTA: 100 cases (US 65%, OUS 35%)
- Automatic Slice Alignment for LV from 2D DE-MRI: 70 cases (US 52%, OUS 48%)
- Standard Initialization of the LV from 2D DE-MRI: 100 cases (US 52%, OUS 48%)
- Standard Initialization of the LV from 3D DE-MRI: 100 cases (US 69%, OUS 31%)
- Standard Initialization of the LA from 3D DE-MRI: 95 cases (US 60%, OUS 35%)
All data in the test set was selected from hospitals not used in any stage of algorithm development (training). The data provenance for the test set includes imaging from both US and OUS (Outside US) countries. The data was anonymized by the hospitals in compliance with GDPR.
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
The ground truth annotations for the test set were generated by two clinical experts. Their qualifications are not explicitly detailed beyond being "clinical experts independent of the clinical experts who established the ground truth of the training dataset." However, the document states the device is intended for use by "qualified medical professionals (cardiologists, electrophysiologists, radiologists or trained technicians) experienced in examining and evaluating cardiovascular MR and CTA images." It can be inferred that these clinical experts possess similar qualifications.
Adjudication method for the test set:
The adjudication method for reconciling differences between the two clinical experts for the test set ground truth is not explicitly stated. It only mentions that ground truth was generated by two independent experts.
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, a multi-reader, multi-case (MRMC) comparative effectiveness study evaluating human readers with and without AI assistance was not described in the provided text. The study focused on the standalone performance of the AI features.
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:
Yes, the performance testing described is a standalone (algorithm only) performance evaluation. The study assessed the accuracy of the machine learning features against expert-generated ground truth, rather than measuring improvements in human reader performance with the device. The goal of the device is stated as "to provide a preliminary initialization of the target structure, which would then be subject to further refinement by the user," indicating a standalone assessment of its initialization capability.
The type of ground truth used:
The ground truth used for the test set was expert consensus / expert annotation. It was generated using the FDA-cleared ADAS 3D software by two independent clinical experts.
The sample size for the training set:
- Standard Initialization of the Left Chambers and Aorta from CTA: 111 DICOM images
- Standard Initialization of the Coronaries from CTA: 231 DICOM images
- Standard Initialization the LA from CTA: 136 DICOM images
- Standard Initialization LV from 2D DE-MRI: 126 DICOM images
- Standard Initialization of the LV from 3D DE-MRI: 110 DICOM images
- Standard Initialization of the LA from 3D DE-MRI: 82 DICOM images
How the ground truth for the training set was established:
The ground truth annotations for the training dataset were "generated initially by the hospitals' clinical teams and revised by Adas3D Medical's Clinical Team." Adas3D Medical's Clinical Team consists of "highly experienced individuals with knowledge of cardiac anatomy, interpretation of MRI and CT volumes, and the use of ADAS 3D."

Summary

Regulation Number and Section

§ 892.2050 Medical image management and processing system.

(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).