Search Results

SwiftMR is a stand-alone medical imaging software solution intended for the acceptance, enhancement, processing, review, analysis, communication, and transfer of all body parts MR images in DICOM format. The software may be used for the enhancement of medical images, such as noise reduction and increased image sharpness for MR images.

The device is designed for use by healthcare professionals and is intended to assist clinicians, who remain responsible for making all final patient management decisions. The device is not intended for use on mobile devices.

The available field strengths are as follows: 0.25T, 0.31T, 0.4T, 0.55T, 0.6T, 1.5T, and 3.0T.

SwiftMR is software used as a Medical Device (SaMD) consisting of a software algorithm that enhances images taken by MRI scanners. The device only processes DICOM images for the end user and is intended to be used by radiology technologists in an imaging center, clinic, or hospital.

The device's inputs are MRI images in DICOM format. The deep learning algorithm produces enhanced images as outputs with reduced noise and increased sharpness in DICOM format. The deep learning algorithm performs noise reduction with the ability of adjusting the denoising level from level 0 to level 8, and sharpening filter performs the sharpening function with the ability of adjusting the sharpness level from level 0 to level 5.

SwiftMR provides an automatic image quality enhancement function for MR images acquired in various environments. SwiftMR can only be used for professional purposes and is not intended for use on mobile devices.

SwiftMR's automation procedure is as follows:

• Receive MR images that are in DICOM format from PACS or from MRI
• Image quality enhancement using Deep Learning model and sharpening filter
• Transfer enhanced MR image as DICOM format to PACS or to MRI

SwiftMR supports input images reconstructed using both conventional vendor reconstruction algorithms and vendor-implemented deep learning (DL) reconstruction pipelines. These vendor DL-reconstructed images are treated as standard DICOM images, and compatibility verification has confirmed that upstream DL processing does not negatively affect SwiftMR performance, artifacts, or anatomical fidelity.

Image Enhance deep leaning model can be applied to MR images with field strengths of 0.25T, 0.31T, 0.4T, 0.55T, 0.6T, 1.5T, and 3.0T. SwiftMR is compatible with both conventional vendor reconstruction methods and vendor-implemented deep learning reconstruction pipelines.

At the same time, SwiftMR allows logged-in users to use its functions and view the processing status. When logged in as the System Admin, the function is available to control automation procedure and system change settings. On the User side, the User can retrieve the results of image processing in the form of a worklist by login to the user account.

The software provides three main functions, which are image processing, quality check and progress monitoring. As part of the image processing functionality, the software performs the following non-deep-learning processing of MR images:

• Diffusion-related processing: ADC, exponential ADC, calculated b-value, fractional anisotropy (FA), FA color, tractography
• Perfusion-related processing: cerebral blood flow, cerebral blood volume, mean transit time, time to peak
• Susceptibility-weighting imaging related processing: filtered phase, phase mask weighting
• 3D-related processing: Maximum Intensity Projection (MIP), Minimum Intensity Projection (minIP), Multi-Planar Reconstruction (MPR)

The software is intended to run automatically in the background so that it does not interrupt the workflow of users. When the user executes MR scans as he/she usually does, the newly acquired images are automatically uploaded to the server and registered in the database (DB) for image processing. Once image processing is complete, the images are sent to PACS or to MR device.

If the user wishes to monitor this automated workflow to check on the status of image processing, he/she can check the main page of the client application or toast messages will appear on the bottom right corner upon completion of each processing. After using the software, they should log out for security reasons.

A settings menu is provided in the form of a user interface to enable the system admin to modify software settings as required by the institution or respective user.

The provided FDA 510(k) clearance letter and summary for SwiftMR contains information about its acceptance criteria and the study conducted to prove it meets these criteria. However, some specific details requested in your prompt (e.g., ground truth for the training set, MRMC study effect size) are not explicitly present in the provided text.

Here's a breakdown of the available information:

1. Table of Acceptance Criteria and Reported Device Performance

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

The document states that "retrospective clinical images" were used for validation testing of both noise reduction and sharpness increase functions.

• Sample Size: The exact number of images or patient cases in the test set is not specified in the provided text. The criteria mention "at least 90% of the dataset" for various performance metrics, indicating a dataset was used, but its size is not quantified.
• Data Provenance: The data consisted of retrospective clinical images. While it mentions the submitter is from the Republic of Korea, the country of origin of the clinical data is not explicitly stated for the test set. The manufacturers of the MRI scanners used are listed (SIEMENS, GE, PHILIPS, CANON, ESAOTE, FONAR, FUJIFILM).

3. Number of Experts Used and Their Qualifications

The document does not specify the number of experts used to establish the ground truth for the test set or their qualifications.

4. Adjudication Method for the Test Set

The document does not specify any adjudication method (e.g., 2+1, 3+1) for establishing ground truth or evaluating the test set results. The performance criteria are quantitative (SNR, FWHM), which suggests an objective, measurement-based evaluation rather than a consensus-based visual assessment for the primary validation.

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

The document does not mention if a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was done to assess how human readers improve with AI vs. without AI assistance. The study focuses on the technical performance of image enhancement (noise reduction and sharpness) rather than reader performance.

6. Standalone (Algorithm Only) Performance

Yes, a standalone performance assessment was conducted. The "Performance data" section explicitly details the verification and validation testing of the SwiftMR software's algorithm, focusing on its ability to enhance images by reducing noise and increasing sharpness. The acceptance criteria for SNR increase and FWHM decrease are directly related to the algorithm's output.

7. Type of Ground Truth Used

The type of ground truth used is primarily quantitative image metrics:

• For noise reduction: Signal-to-Noise Ratio (SNR). This usually implies a comparison against a "noise-free" or "reference" image, or a statistical calculation based on signal and noise characteristics within the image itself.
• For sharpness increase: Full Width at Half Maximum (FWHM) of a selected Region of Interest (ROI). This is a direct measurement of image resolution or sharpness.

The document does not explicitly state that expert consensus, pathology, or outcomes data were used as ground truth for these specific performance metrics.

8. Sample Size for the Training Set

The document does not specify the sample size used for the training set. It only mentions that the device uses a "deep learning algorithm."

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

The document does not specify how the ground truth for the training set was established. Given the nature of image enhancement (noise reduction and sharpness), training data for such models often involve "paired" noisy/sharp and "clean/reference" images generated through simulation, high-resolution acquisitions, or expert-curated "ideal" images. However, this is not detailed in the provided text.

Ask a specific question about this device