(105 days)
SyMRI is a post-processing software medical device intended for use in visualization of soft tissue. SyMRI analyzes input data from MR imaging systems. SyMRI utilizes data from supported MR sequences to generate parametric maps of R1, R2 relaxation rates, and proton density (PD).
SyMRI is intended for automatic labeling, visualization and volumetric quantification of segmentable brain tissues from a set of MR images. Brain tissue volumes are determined based on modeling of parametric maps from SyMRI.
SyMRI can also generate multiple image contrasts from the parametric maps. SyMRI enables post-acquisition image contrast adjustment.
SyMRI is indicated for head imaging.
When interpreted by a trained physician, output from SyMRI can provide information useful in determining diagnosis. SyMRI 2D is intended to be used in combination with at least one other, conventional MR acquisition (e.g. T2-FLAIR). T1W and T2W images from SyMRI 3D may replace conventional MR images in a clinical setting when interpreting together with a conventional 3D T2W FLAIR image.
SyMRI works by post-processing a multi-delay, multi-echo acquisition into parametric maps. The acquisition is either a multi-slice 2D approach (M2D-MDME), consisting of 4 delays with a short and a long echo time each (8 images per slice), or a 3D approach (3D-QALAS) consisting of 4 delays with a short echo and 1 delay with a long echo time (5 images per slice).
The parametric maps are R1, R2 relaxation rates, and proton density (PD). The inverse relaxation parameters, T1 relaxation time (1/R1), and T2 relaxation time (1/R2) are also provided as parametric maps.
SyMRI also enables the users to obtain volumetric information in the head, including white matter (WM), gray matter (GM), cerebrospinal fluid (CSF), Myelin correlated (MyC) partial volume, brain parenchyma (BP) and intracranial cavity (IC). This is accomplished by using tissue definitions based on the parametric maps. The tissue definitions provide tissue partial volume, or tissue fraction, per voxel. SyMRI also provides image processing tools to extract the values of the parametric maps, and tissue partial volume, per individual voxel, per region of interest, or the entire imaging volume.
The parametric maps can be visualized as contrast weighted MR images, such as T1, T2, PD, and Inversion Recovery (IR) weighted images (including T1-FLAIR, STIR, Double IR, and PSIR weighted images).
The parametric maps can be visualized as contrast weighted MR images from SyMRI 3D may replace conventional MR images in a clinical setting when interpreting together with a conventional 3D T2W FLAIR image.
SyMRI calculates the pixel signal intensity as a function of R1, R2, PD, and desired MR scanner settings, such as echo time (TE), repetition time (TR), and inversion delay time (TI). A number of default settings for TE, TR, and TI are provided, but the user has the ability to change the contrast of the images. SyMRI generates all the different image contrasts from the same parametric maps, derived from the same acquisition. This leads to enhanced image slice registration, owing to the absence of inter-acquisition patient movement. SyMRI provides the user the ability to change the contrast of the images after the acquisition. This is performed by adjusting the TE, TR, and/or TI parameters post-acquisition, to generate the specific contrast desired.
SyMRI is intended to be used on data produced by any of the following acquisition sequences:
- . MDME sequence data from GE MAGiC
- MDME sequence data from Philips SyntAc
- . MDME sequence data from Siemens TSE_MDME
- 3D-QALAS sequence data from Philips 3DsyntAc
This document describes the acceptance criteria and study proving the device meets them for SyMRI.
1. Table of Acceptance Criteria and Reported Device Performance
| Performance Metric | Acceptance Criteria | Reported Device Performance |
|---|---|---|
| Quantitative Accuracy & Precision (R1, R2, PD) | Equivalent to predicate device. Correspondence with reference values (gold standard phantoms). | R1, R2, and PD measurements show correspondence with reference values (inversion recovery for R1, CPMG multi-echo for R2, heavy water phantoms for PD, and NIST/ISMRM Model 130 phantom). The subject device met the same predefined acceptance criteria as the predicate device, demonstrating equivalence in accuracy and precision for quantification when compared to gold standards. |
| Segmentation Accuracy & Precision | Equivalent to predicate device. | The verification results demonstrate that the subject device SyMRI meets the same pre-defined performance criteria as the predicate. |
| Non-inferiority of Synthetic 3D Images (Diagnostic Performance) | Non-inferiority in sensitivity and specificity for detecting any pathology compared to conventional 3D images. Non-inferiority in diagnostic accuracy of radiological finding class compared to conventional 3D images. | Synthetic 3D images were non-inferior in terms of sensitivity and specificity in detecting any pathology, as well as non-inferior in diagnostic accuracy of radiological finding class, compared to equivalent conventional MR images over a wide range of brain pathologies. |
| Legibility of Anatomical Structures | High legibility of anatomical structures. | All images (synthetic and conventional) had a very high legibility of anatomical structures. |
| Artifact Prevalence | Lower prevalence of artifacts in synthetic images compared to conventional images. No novel artifacts. | Synthetic images had lower prevalence of artifacts compared to the conventional MR images. No novel artifacts were reported for synthetic MR images. |
| Image Quality Score | Higher image quality score for synthetic images compared to conventional images. | Synthetic images had slightly higher image quality scores compared to conventional images for both T1W and T2W images. |
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size: 189 subjects
- Data Provenance:
- Country of Origin: United States (6 institutes in the US)
- Study Type: Prospective, multi-reader clinical investigation.
- Subject characteristics: Patients with a wide range of different pathologies, and healthy controls. Both adults and pediatric patients were included.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Their Qualifications
The document does not explicitly state the number of experts used to establish a ground truth for the test set. However, it mentions that five experienced radiologists assessed the images in the clinical investigation. Their qualifications are described as "experienced radiologists."
4. Adjudication Method for the Test Set
The document does not explicitly state an adjudication method (e.g., 2+1, 3+1). It states that "Five experienced radiologists... assessed the images in two reading sessions with a four-week memory washout period in between." This suggests individual assessments rather than a consensus-based adjudication for primary readings, though further details on how discrepancies (if any) were handled are not provided.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- Yes, a MRMC comparative effectiveness study was done. The study involved five experienced radiologists assessing images in two reading sessions.
- Effect Size of Human Reader Improvement with AI vs. Without AI Assistance: This specific information (effect size of human readers improving with AI assistance vs. without) is not directly provided. The study focused on demonstrating the non-inferiority of synthetic 3D images (generated by the AI device, SyMRI) compared to conventional 3D images in terms of diagnostic performance (sensitivity, specificity, diagnostic accuracy). It also evaluated image quality and artifact prevalence of the synthetic images. It does not describe a scenario where human readers interpreted conventional images and then re-interpreted them with AI assistance to measure improvement. Instead, it compares the diagnostic utility of AI-generated images versus conventional images.
6. Standalone (Algorithm Only) Performance Study
Yes, a standalone performance assessment was conducted for the quantitative aspects of the device:
- Accuracy of R1/R2/PD quantification: Evaluated compared to "gold standard inversion recovery (R1), CPMG multi-echo (R2), heavy water phantoms (PD) and standard system Model 130 NIST/ISMRM phantom." This demonstrates an algorithm-only accuracy assessment against established physical standards.
- Segmentation Accuracy & Precision: Verified against pre-defined performance criteria similar to the predicate device. While not explicitly stated as "algorithm only," the context of "verification results" for quantification and segmentation usually refers to algorithmic performance.
7. Type of Ground Truth Used
- For Quantitative Parameters (R1, R2, PD): Gold standard phantoms (inversion recovery, CPMG multi-echo, heavy water phantoms, NIST/ISMRM Model 130 phantom).
- For Diagnostic Performance (Pathology Detection, Diagnostic Accuracy, Image Quality): The consensus or individual expert assessments of the five experienced radiologists after reviewing both synthetic and conventional images served as the reference for comparison, aiming to establish non-inferiority against conventional imaging. Although not explicitly called "ground truth," the conventional images and expert interpretations thereof served as the benchmark. The study compared the device-generated synthetic images to equivalent conventional MR images. It also mentions "radiological finding class," implying expert-derived classifications.
8. Sample Size for the Training Set
The document does not provide information on the sample size used for the training set of the SyMRI algorithm. It focuses on the validation study.
9. How the Ground Truth for the Training Set Was Established
The document does not provide information on how the ground truth for the training set was established, as it does not describe the training process for the algorithm. It primarily details the performance validation of the device.
§ 892.1000 Magnetic resonance diagnostic device.
(a)
Identification. A magnetic resonance diagnostic device is intended for general diagnostic use to present images which reflect the spatial distribution and/or magnetic resonance spectra which reflect frequency and distribution of nuclei exhibiting nuclear magnetic resonance. Other physical parameters derived from the images and/or spectra may also be produced. The device includes hydrogen-1 (proton) imaging, sodium-23 imaging, hydrogen-1 spectroscopy, phosphorus-31 spectroscopy, and chemical shift imaging (preserving simultaneous frequency and spatial information).(b)
Classification. Class II (special controls). A magnetic resonance imaging disposable kit intended for use with a magnetic resonance diagnostic device only is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 892.9.