Search Results

The MAGNETOM system is indicated for use as a magnetic device (MRDD) that produces transverse, sagittal, coronal and oblique cross sectional images, and that displays the internal structure and/or function of the head or extremities. Other physical parameters derived from the images may also be produced. Additionally, the MAGNETOM system is intended to produce Sodium images for the head and Phosphorus spectroscopic images and/or spectra for whole body, excluding the head. These images and/or spectra and the physical parameters derived from the images and/or spectra when interpreted by a trained physician yield information that may assist in diagnosis.

MAGNETOM Terra and MAGNETOM Terra.X with software syngo MR XA60A include new and modified hardware and software compared to the predicate device, MAGNETOM Terra with software syngo MR E12U. A high level summary of the new and modified hardware and software is provided below: Hardware: New Hardware (Combiner (pTx to sTx), MC-PALI, GSSU control unit, 8Tx32Rx Head coil), Modified Hardware (Main components such as: Upgrade of GPA, New Host computer hardware, New MaRS computer hardware, Upgrade the SEP, The new shim cabinet ASC5 replaces two ACS4 shim cabinets; Other components such as: RFPA, Use of a common MR component which provides basic functionality that is required for all MAGNETOM system types, The multi-nuclear (MNO) option has been modified, OPS module, Cover with UI update on PDD). Software: New Features and Applications (Static B1 shimming, TrueForm (1ch compatibility mode), Deep Resolve Boost, Deep Resolve Gain, Deep Resolve Sharp, Bias field correction (marketing name: Deep RxE), The new BEAT pulse sequence type, BLADE diffusion, The PETRA pulse sequence type, TSE DIXON, The Compressed Sensing (CS) functionality is now available for the SPACE pulse sequence type, The Compressed Sensing (CS) functionality is now available for the TFL pulse sequence type, IDEA, The Scientific Suite), Modified Features and Applications (EP2D DIFF and TSE with SliceAdjust, The Turbo Flash (TFL)), Modified Software / Platform (Stimulation monitoring, "dynamic research labeling"), Other Modifications and / or Minor Changes (Intended use, SAR Calculation and Weight limit reduction for 31P/1H TxRx Flex Loop Coil, X-upgrade for MAGNETOM Terra to MAGNETOM Terra.X, Provide secure MR scanner setup for DoD (Department of Defense) -Information Assurance compliance).

The provided text describes the acceptance criteria and supporting study for the AI features (Deep Resolve Boost, Deep Resolve Sharp, and Deep RxE) within the MAGNETOM Terra and MAGNETOM Terra.X devices.

Here's a breakdown of the requested information:

1. Table of Acceptance Criteria and Reported Device Performance

Note on Acceptance Criteria: The document directly states acceptance criteria for Deep RxE (e.g., NRMSE < 11%). For Deep Resolve Boost and Deep Resolve Sharp, the "acceptance criteria" are more implicitly derived from the described validation and evaluation metrics and outcomes (e.g., "successful passing of quality metrics tests," "increased edge sharpness," "no misinterpretation").

2. Sample Size Used for the Test Set and Data Provenance

Patient characteristics (Gender/Age) were recorded for Deep RxE: female: 56%, male: 41%, phantom: 3%. Age range 20-80 years. Not recorded for other features.
Ethnicity was not recorded for any feature. The document states that due to network architecture, attributes like gender, age, and ethnicity are not relevant to training data.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications

• Deep Resolve Boost & Deep Resolve Sharp: The document does not mention the use of experts for ground truth establishment for the test set regarding these features. Images were visually inspected and quality metrics were used.
• Deep RxE: The document mentions that images acquired with Deep RxE were "rated better for image quality than the ones acquired without it in the clinical study that was conducted." This implies expert evaluation, but the number of experts or their qualifications for the test set ground truth for Deep RxE is not explicitly stated.
  • Separately, for the overall device clearance, "radiologist's evaluation reports from two U.S. board-certified radiologists have been provided" for software modifications and new hardware. This is a general statement for the device and not specifically linked to the ground truth of the AI features' test sets.

4. Adjudication Method for the Test Set

The document does not explicitly describe an adjudication method (like 2+1, 3+1) for establishing ground truth for the test sets of these AI features. For Deep Resolve Boost and Sharp, the ground truth was derived from the acquired datasets themselves, which were then manipulated to create input data. For Deep RxE, the ground truth for phantom testing was a "previously defined reference output" or the acquired data, and for the clinical study, images were "rated better" but the adjudication process for this rating isn't detailed.

5. Multi Reader Multi Case (MRMC) Comparative Effectiveness Study

No MRMC comparative effectiveness study is directly mentioned specifically for the AI features (Deep Resolve Boost, Deep Resolve Sharp, or Deep RxE) that compares human readers with vs. without AI assistance. The document alludes to radiologists evaluating images with new software features or comparing images from subject/predicate devices, and for Deep Resolve Boost, it mentions clinical settings and "seven peer-reviewed publications" concluding beneficial use for clinical routine, with reports of "significant time savings." For Deep RxE, images were "rated better for image quality," which implies a reader study, but no details on methodology, number of readers, or specific effect size are provided to quantify human reader improvement with AI assistance.

6. Standalone (Algorithm Only) Performance

Yes, standalone performance was conducted for all three AI features:

• Deep Resolve Boost: Characterized by PSNR and SSIM, and visual inspection.
• Deep Resolve Sharp: Characterized by PSNR, SSIM, perceptual loss, visual rating, and intensity profile comparisons.
• Deep RxE: Performance of the network tested on a phantom against ground truth (maximal allowed NRMSE of 11% for 2D, 8.7% for 3D achieved). Unit-tests and a two-step test procedure involving validation on unseen data and RMS error calculation against ground truth.

These indicate evaluation of the algorithm's performance without a human in the loop, beyond initial visual inspections by evaluators.

7. Type of Ground Truth Used

• Deep Resolve Boost: The acquired datasets themselves, retrospectively manipulated through data manipulation and augmentation (under-sampling, lowering SNR, mirroring k-space data) to create input data.
• Deep Resolve Sharp: The acquired datasets themselves, retrospectively manipulated through data manipulation (cropping k-space data) to create corresponding low-resolution input and high-resolution output/ground truth.
• Deep RxE: For phantom testing, a "previously defined reference output" was used, and for other evaluations, the acquired datasets themselves were used for comparison against bias field correction methods (homodyne filtering, N4, UNICORN).

8. Sample Size for the Training Set

9. How the Ground Truth for the Training Set Was Established

• Deep Resolve Boost: The "acquired datasets represent the ground truth for the training and validation." Input data was "retrospectively created from the ground truth by data manipulation and augmentation."
• Deep Resolve Sharp: The "acquired datasets represent the ground truth for the training and validation." Input data was "retrospectively created from the ground truth by data manipulation."
• Deep RxE: The document states that "During training the loss, as the difference to a ground truth, is monitored." The method of establishing this initial ground truth is implicitly the raw acquired data from the 7T MRI scanners, as the network aims to correct for B1 inhomogeneities. It also states "All data from the two MR systems were separated into independent training, validation and test datasets," implying the raw or processed raw data served as reference for training.

Ask a specific question about this device