(24 days)
The nordicMEDiVA software is an advanced visualization and processing platform with a specific focus on providing algorithms designed to analyze functional and dynamic MRI data of the brain. The software runs on a server in a networked environment and is accessed by users via a standard web browser. It can communicate with other imaging platforms that support DICOM, and process medical image data acquired through DICOM-compliant imaging devices and modalities.
nordicMEDiVA is indicated for image analysis and visualization of functional and dynamic MRI data of the brain, presenting derived properties and parameters from the input image data in a clinically useful context.
nordicMEDIVA is a software as a medical device (SaMD) for processing of MR images of the brain. Users will configure analysis pipelines, which are executed automatically when image data is received or manually by a user. The user can choose to send the results to other DICOM nodes for review or use nordicView for their review and export the results to PACS, neuro navigation systems, or other DICOM-compliant modalities.
nordicMEDIVA is a server-client solution and can be installed on a local server at the customer's location or in a cloud-based setup. The software is containerized with Docker technology and operates on a GPU-enabled Linux host. This allows customers to manage the server environment themselves or use it as a Service (SaaS) hosted by NordicimagingLab AS in the cloud. Customers can install the server on physical hardware, virtual machines, or in their own cloud infrastructure.
The device comprises a database, DICOM functionality, various APIs, a visualization engine, and medical image analysis modules. The device is not intended for long-term persistent storage of medical diagnostic data.
The device incorporates rule-based algorithms for the calculation of metrics from dynamic MRI data. The device does not incorporate AI algorithms based on neural networks.
The device connects to other imaging modalities, such as MR scanners, PACS, and surgical navigation systems.
The following modules provide the main functions of the device.
nordicView: A browser-based user interface accessed from desktop clients that provides tools for general image visualization, export, and relevant analysis tools for BOLD-fMRI and DSC-perfusion.
nordicBOLD: BOLD task-based fMRI anall magnetic susceptibility changes in the human brain in areas with altered blood flow resulting from neuronal activity. The image processing requires the definition of a so-called design matrix which is used to calculate voxel-wise statistics conveying information about the probability of the execution of the qiven task.
The design matrix is is defined such that the task or stimulation that was presented to the patient during the scan. The task or stimulation presented during scan time is often refered to as "the paradigm". The design matrix can be defined manually by the user, or a paradigm from nordicAktiva - another product from NordicNeuroLab - can be used.
nordicAktiva is a software, marketed by NordicNeuroLab, that may be used during scan time to present the patient or subject being scanned. The use of nordicAktiva is not required.
nordicDSC: Calculations of perfusion-related parameters that provide information about the blood vessel structure and characteristics. Such maps include blood volume, blood flow, time to peak, mean transit time, and leakage.
Platform: The platform includes a database, DICOM functionality, various APls, processing pipelines for the medical image analysis modules. Serves as a backbone component for the other modules of nordicMEDiVA.
Dashboard: a browser-based user interface accessed from desktop clients for administration and configuration.
The provided text describes the acceptance criteria and the study that proves the device meets those criteria for the nordicMEDiVA software.
Here's the breakdown:
1. Table of Acceptance Criteria and Reported Device Performance
| Module/Test | Acceptance Criteria | Reported Device Performance |
|---|---|---|
| Usability | All scenarios from summative usability test meet acceptance criteria. | All scenarios from the summative usability test met the acceptance criteria completely; no new risks were found, and existing risk control measures were proven to be effective. |
| nordicBOLD | Results proven to be as effective as the predicate device (nordicBrainEx). | The results from nordicBOLD were evaluated in comparison with equivalent results from the predicate device, nordicBrainEx (K163324). The results were reviewed by internal and external clinical experts and proven to be as effective as the predicate device. |
| nordicDSC | Lin's Concordance Correlation Coefficient (CCC) of enhancing voxels >= 0.8. | The results from the nordicDSC were confirmed to be the same as the predicate (nordicDSC K212720) where a Lin's Concordance Correlation Coefficient of enhancing voxels was calculated with an acceptance criterion being greater than or equal to 0.8 was applied. (Note: Specific CCC value achieved is not explicitly stated, only that it met the criteria). |
| Cybersecurity | Conforms to cybersecurity requirements. | nordicMEDIVA conforms to cybersecurity requirements by implementing a means to prevent unauthorized access, modification, misuse, denial of use of information stored, accessed or transferred from a medical device to an external recipient. |
2. Sample Size Used for the Test Set and Data Provenance
The document does not explicitly state the numerical sample size for the test set. For nordicBOLD and nordicDSC, the comparison was made against existing predicate devices with existing data. The data provenance (country of origin, retrospective/prospective) is also not specified beyond the fact that it was compared against previously cleared devices.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications
- nordicBOLD: "internal and external clinical experts" were used to review the results. The exact number and specific qualifications (e.g., "radiologist with 10 years of experience") are not specified.
- nordicDSC: The ground truth for nordicDSC seems to be based on direct comparison to the predicate device (nordicDSC K212720) through a statistical measure (Lin's CCC), rather than expert adjudication of a new test set.
4. Adjudication Method for the Test Set
- nordicBOLD: Reviewed by "internal and external clinical experts." While experts were involved, a specific adjudication method (e.g., 2+1, 3+1 consensus) is not explicitly stated. It implies a qualitative "proven to be as effective" rather than a strict quantitative adjudication.
- nordicDSC: Adjudication method is not applicable in the traditional sense, as it was a quantitative comparison using Lin's CCC against a predicate.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was Done
No, a multi-reader multi-case (MRMC) comparative effectiveness study explicitly designed to show how much human readers improve with AI vs. without AI assistance was not described in the provided text. The evaluation focused on the performance of the software (nordicBOLD and nordicDSC) in comparison to predicate devices, and usability, not on human reader performance with AI assistance. The device does not incorporate AI algorithms based on neural networks, limiting the scope for typical AI-assisted MRMC studies.
6. If a Standalone (i.e. algorithm only without human-in-the-loop performance) was done
Yes, from the descriptions of the nordicBOLD and nordicDSC evaluations, they appear to be standalone performance assessments of the algorithms' output compared to predicate devices. The "results from nordicBOLD were evaluated in comparison with equivalent results from the predicate device" and "results from the nordicDSC were confirmed to be the same as the predicate" indicates an algorithm-only evaluation, followed by expert review for nordicBOLD.
7. The Type of Ground Truth Used
- nordicBOLD: The ground truth appears to be established by comparison to the results generated by the predicate device (nordicBrainEx) and reviewed by "internal and external clinical experts." This suggests a comparative ground truth based on established clinical performance of a cleared device, with expert consensus on equivalency.
- nordicDSC: The ground truth for comparison was the output of the predicate device (nordicDSC K212720), with equivalency quantified by Lin's Concordance Correlation Coefficient.
8. The Sample Size for the Training Set
The document does not specify the sample size for the training set. It states that the device does not incorporate "AI algorithms based on neural networks," implying that traditional machine learning (which often requires training data) or deep learning was not the primary methodology. The device uses "rule-based algorithms for the calculation of metrics from dynamic MRI data."
9. How the Ground Truth for the Training Set was Established
Given that the device "does not incorporate AI algorithms based on neural networks" and uses "rule-based algorithms," the concept of a "training set" with established ground truth as typically understood for deep learning models is likely not applicable. The algorithms are rule-based, meaning their performance depends on the pre-defined rules, which are likely derived from scientific principles, clinical knowledge, and established methodologies for fMRI and DSC analysis, rather than learned from a labeled training dataset.
§ 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).