(72 days)
Ezra Flash is an image processing software used for image enhancement of MR images. It can be used to reduce image noise in images acquired as part of non-contrast MRI exams on 3-Tesla Siemens for Sagittal T1, Axial T2 and Axial Flair sequences within the head region for patients > 18 years of age.
Ezra Flash is a Software as a Medical Device (SaMD) consisting of a software algorithm that enhances images of the head region taken by MRI scanners. As it only processes images for the end user, the device has no interface. It is intended to be used by radiologists in an imaging center, clinic, or hospital. The software can be used with MR images acquired as part of MRI exams on 3-Tesla Siemens and GE scanners. The outputs are images with enhanced image quality. Both the original non-enhanced studies and the Ezra Flash-enhanced studies are available to the end user. Ezra Flash receives DICOM-compliant non-contrast MR image inputs acquired on 3-Tesla scanners for Sagittal T1, Axial T2 and Axial Flair sequences within the head region. The software uses a convolutional neural network-based algorithm to improve image quality by reducing noise. The device outputs a DICOM-compliant copy of the images with improved image quality.
Ezra Flash Acceptance Criteria and Performance Study
The Ezra Flash device for image enhancement of MR images (K230264) underwent performance testing to demonstrate its safety and effectiveness. The study focused on quantitative metrics (SNR, CNR) and a subjective image quality assessment by experts (Perceived Noise).
1. Table of Acceptance Criteria and Reported Device Performance
| Metric | Acceptance Criteria | Reported Device Performance |
|---|---|---|
| Signal-to-Noise Ratio (SNR) | SNR of a selected region of interest (ROI) in each test dataset is on average improved by > 5% after Ezra Flash enhancement compared to original MR-acquired images (raw). | The test results demonstrated that the Ezra Flash performs to its intended use and is deemed acceptable for clinical use. (Specific numerical improvement not provided, but implies criterion was met). |
| Contrast-to-Noise Ratio (CNR) | CNR of a selected region of interest (ROI) in each test dataset is on average improved by > 0% after Ezra Flash enhancement compared to the MR-acquired raw images. | The test results demonstrated that the Ezra Flash performs to its intended use and is deemed acceptable for clinical use. (Specific numerical improvement not provided, but implies criterion was met). |
| Image Quality Perceived Noise | The mean Likert results for the Ezra Flash-enhanced images compared to the original MR-acquired images (raw) is greater than or equal to 0.5 Likert scale points. | The test results demonstrated that the Ezra Flash performs to its intended use and is deemed acceptable for clinical use. (Specific numerical improvement not provided, but implies criterion was met). |
2. Sample Size Used for the Test Set and Data Provenance
The document does not explicitly state the sample size for the test set or the data provenance (e.g., country of origin, retrospective or prospective). It mentions "each test dataset" for SNR and CNR, suggesting multiple datasets were used.
3. Number of Experts Used to Establish Ground Truth for the Test Set and Qualifications
The document mentions "Likert results for the Ezra Flash-enhanced images compared to the original MR-acquired images (raw)," which implies human expert assessment of image quality, likely for perceived noise. However, it does not specify:
- The number of experts used.
- The qualifications of those experts (e.g., specific years of experience, board certification).
4. Adjudication Method for the Test Set
The document does not describe any specific adjudication method (e.g., 2+1, 3+1) for the expert ratings on the test set. It only refers to "mean Likert results."
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
The provided text does not indicate that a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was done. The performance testing description focuses on objective image metrics (SNR, CNR) and a subjective image quality assessment, but not on human reader performance with and without AI assistance. Therefore, no effect size for human reader improvement is provided.
6. Standalone (Algorithm Only) Performance Study
Yes, a standalone performance study was done for the Ezra Flash algorithm. The described testing focuses on the algorithm's direct impact on image quality metrics (SNR, CNR, perceived noise) by processing raw MR images and comparing them to the enhanced outputs. The device "only processes images for the end user" and "has no interface," with outputs viewed on existing PACS workstations. This indicates evaluation of the algorithm's performance in isolation from a human-in-the-loop workflow.
7. The Type of Ground Truth Used
The ground truth for the quantitative metrics (SNR, CNR) is inherently derived from the raw MR-acquired images themselves, serving as the baseline for improvement. For the "Image Quality Perceived Noise" metric, the ground truth is based on expert consensus (or at least expert scores) derived from Likert scale ratings comparing enhanced images to raw images. This is a form of subjective expert evaluation.
8. The Sample Size for the Training Set
The document does not specify the sample size for the training set. It mentions that the device uses a "convolutional neural network-based algorithm" whose "parameters of the filters were obtained through an image-guided optimization process," implying a training phase. However, the specific number of images or cases used for training is not disclosed.
9. How the Ground Truth for the Training Set was Established
The document states that the "parameters of the filters were obtained through an image-guided optimization process." While this indicates that the model was trained, it does not explicitly describe how the ground truth for the training set was established. Typically, for image enhancement, training ground truth might involve:
- Paired low-quality/high-quality images.
- Synthetic noise addition to high-quality images.
- Expert-annotated "ideal" images or metrics.
However, the provided text does not elaborate on this aspect.
§ 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).