Search Results

Vitrea® Image Denoising software is intended to assist radiologists and specialists in the enhancement of CT and 3D-XA image presentation by enabling noise reduction and contrast enhancement technique.

The Vitrea Image Denoising software is a tool available on the Vitrea platform. It assists radiologists and specialists in the enhancement and viewing of CT and 3D-XA images from a variety of diagnosis imaging systems by noise reduction and contrast enhancement. Vitrea Image Denoising software employs 3D analysis of the image structure of each voxel. Random noise does not have a 3D structure and can thus be separated from dominant structures. Once the structure has been determined, the denoising tool suppresses noise by averaging voxel information without removing important structural details for reducing the noise.

It provides a control to turn on or off a denoising preset in Multi-planar Reformatting (MPR) and 3D views to reduce noise and enhance contrast in reconstructed CT and 3D-XA image datasets, while preserving structural details of contrast, spatial size, and 3D structure in the native images for visual assessment. The user can interactively apply and remove the noise reduction filter during the image review to assess the effect of the noise reduction on the images and generate a new series of denoised snapshots for further review in other workstations or PACS. The denoised images can be used in conjunction with the original images as the user can switch between the original image and the denoised image while conducting image review. It also enables the user to create new denoising presets, edit existing denoising presets, and save the current denoising values as a custom preset.

Vitrea® Image Denoising Software (K140395) - Acceptance Criteria and Study Summary

This device is a software tool intended for image enhancement via noise reduction and contrast enhancement in CT and 3D-XA images. It does not diagnose or manage disease, but rather improves the visual quality of existing images for radiologists and specialists.

1. Table of Acceptance Criteria and Reported Device Performance

The provided document does not explicitly state quantitative acceptance criteria for device performance in terms of specific metrics like Signal-to-Noise Ratio (SNR) improvement, resolution maintenance, or contrast enhancement percentages. Instead, the acceptance criteria are framed qualitatively around the preservation of diagnostic information and the perception of noise reduction by experts.

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

• Sample Size for Test Set: The document does not specify a numerical sample size of medical images used for the external validation (usability testing). It states "various Catphan phantoms" were used for internal validation, and "previously acquired medical images" were used for software testing.
• Data Provenance:
  • Internal Validation (Phantom Testing): Catphan phantoms (likely synthetically generated or physical phantoms used in a controlled environment).
  • External Validation: "Previously acquired medical images" were used. No specific country of origin is mentioned, but the review was conducted by US-based radiologists and an interventional cardiologist. The data is retrospective as it refers to "previously acquired" images.

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

• Number of Experts: "Experienced Radiologists and an Interventional Cardiologist" were used for external validation. The exact number is not explicitly stated, but the phrasing suggests more than one radiologist.
• Qualifications of Experts: They are described as "experienced Radiologists and an Interventional Cardiologist." Specific years of experience are not provided.

4. Adjudication Method for the Test Set

The document does not describe a formal adjudication method (e.g., 2+1, 3+1). The external validation involved experts evaluating the software and providing feedback ("Each user felt that..."). This suggests individual assessment rather than a consensus-driven adjudication process for ground truth establishment. Since the device enhances images for visual assessment, subjective expert opinion on image quality is a primary measure.

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

No MRMC comparative effectiveness study is mentioned, nor is an effect size indicating how much human readers improve with AI vs. without AI assistance. The study focuses on evaluating the standalone performance of the denoising software and its perceived impact on image quality.

6. Standalone (Algorithm Only Without Human-in-the-Loop) Performance

Yes, a standalone performance assessment was conducted implicitly through the internal validation using phantom testing. This measured the impact of the software on noise reduction, spatial resolution, sharpness, and low contrast resolution in a controlled environment. The external validation also assesses the algorithm's output directly, as users "assess the effect of the noise reduction on the images."

7. Type of Ground Truth Used

• Internal Validation: Defined characteristics of phantom images were used as ground truth (e.g., expected noise levels, spatial resolution, low contrast details).
• External Validation: The subjective opinion and visual assessment of experienced medical professionals (Radiologists and an Interventional Cardiologist) served as the ground truth for evaluating the clinical relevance and effectiveness of the image enhancement. There is no mention of pathology or outcomes data being used as ground truth for this device.

8. Sample Size for the Training Set

The document does not provide information on the sample size for the training set. The device description focuses on its algorithmic approach (3D analysis of image structure, separating random noise) rather than explicitly mentioning machine learning or a training phase specific to a dataset.

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

Given the absence of information regarding a training set and the description of the denoising method as an algorithmic approach (3D analysis of image structure), it is not explicitly stated how ground truth for a training set was established. It's possible the algorithm relies on inherent image properties and mathematical models rather than a supervised learning approach with labeled training data in the traditional sense of AI/ML.

Ask a specific question about this device