Search Results

Kinepict Medical Imaging Tool version 4.0 (KMIT v4.0) is intended to visualize blood vessel structures by detecting the movement of the contrast bolus during angiography examination and to give the medical experts the option to reduce x-ray dose. The effectively achievable radiation reduction depends on the characteristics of the different cath labs (angiography unit, acquisition protocol, anatomical region) and requires the optimization of locally appled acquisition protocols. The software is intended to be used in addition to, or as a replacement for current DSA imaging. KMT v4.0 can be deployed on independent workstation hardware for stand-alone diagnostic assessment, post-processing, reporting. It can be configured within a network to send and receive DICOM data. Furthermore, KMIT v4.0 can be deployed on several kinds of angiography systems. It provides solutions in the operating room for imageguided solutions for interventional oncology, interventional radiology, and interventional neuroradiology. KMIT v4.0 can be also combined with fluoroscopic or radiographic systems.

The Kinepict Medical Imaging Tool version 4.0 (KMIT v4.0) is medical diagnostic software for real-time viewing, diagnostic review, post-processing, optimization, communication, reporting, and storage of medical images and data on exchange media. It provides imageguided solutions in the operating room, for image-guided surgery, and image-guided solutions for interventional oncology, interventional radiology, and interventional neuroradiology. KMIT v4.0 can be deployed on independent workstation hardware for stand-alone diagnostic assessment, post-processing, reporting, which are intended to assist the physician in the evaluation of digital radiographic examinations, including diagnosis and/or treatment planning. KMIT v4.0 is designed to work with digital radiographic, fluoroscopic, interventional, and angiographic systems. The algorithm behind the Digital Variance Angiography (DVA) image calculation is the same in KMIT v2.2 (Kinepict Health Ltd. under K190993) and v4.0 (see Device Description / Calculation of DVA section). The postprocessing functions like contrast and brightness settings, choosing mask image, pixel shift applications, and anonymizing options are the same in the two software. Image storing and image sending functions are using the same DICOM technic and ports as Kinepict Medical Imaging Tool v2.2. In v4.0 automatic upload to PACS and automatic opening of screen size view is implemented. These differences do not have any effect on safety and efficiency. In summary, the KMIT v4.0 software does not introduce any new potential safety risks and is substantially equivalent to and performs as well as the predicate device.

Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided text:

Kinepict Medical Imaging Tool 4.0 - Acceptance Criteria and Study Details

The primary acceptance criteria for the Kinepict Medical Imaging Tool 4.0 (KMIT v4.0) revolve around its ability to maintain diagnostic quality while enabling a reduction in X-ray radiation dose during angiography. The key performance metrics evaluated are:

• Non-inferiority of low-dose DVA (ldDVA) images compared to normal dose DSA (ndDSA) images in terms of visual evaluation scores.
• No significant difference in the number of recognized arteries and the proportion of arteries suitable for diagnosis between ldDVA and ndDSA.
• High sensitivity and specificity of ldDVA in reproducing diagnostic categories established by ndDSA.

1. Table of Acceptance Criteria and Reported Device Performance

2. Sample Size and Data Provenance

• Sample Size (Test Set): 30 Peripheral Artery Disease (PAD) patients.
• Data Provenance: Prospective interventional clinical study conducted between April and July 2019. The country of origin of the data is not explicitly stated, but the company is based in Hungary. The study was performed using a Siemens Artis Zee angiography system and Ultravist 370 iodinated contrast medium.

3. Number of Experts and Qualifications for Ground Truth

• Visual Evaluation: Seven specialists evaluated image quality using a 5-grade Likert scale. Their specific qualifications (e.g., years of experience, subspecialty) are not explicitly stated, beyond being referred to as "specialists" in the field of angiography.
• Task-Based Survey: Six readers identified artery segments and evaluated stenosis. Their specific qualifications are not explicitly stated.
• Discordant Decisions Adjudication: An "expert" determined the valid diagnostic category when decisions were discordant. The number of such experts and their specific qualifications are not explicitly stated beyond "an expert."

4. Adjudication Method for the Test Set

• Visual Evaluation: No adjudication method is explicitly stated for the 7 specialists' Likert scale evaluations. The text suggests that the scores from these specialists were analyzed directly (e.g., "ldDVA received significantly higher visual evaluation scores...").
• Task-Based Diagnostic Test: For the 6 readers, no adjudication method is explicitly stated for their initial identification of arteries or stenosis evaluation.
• Discordant Decisions: When the 6 readers had "discordant decisions" regarding diagnostic categories, an "expert" supervised and determined the valid diagnostic category. This suggests a form of adjudication for disagreements, but the specific process (e.g., 2+1, 3+1) is not detailed beyond the involvement of a single expert for supervision.

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

• Yes, a form of MRMC study was conducted. The study involved 7 "specialists" for visual evaluation and 6 "readers" for a task-based diagnostic test, evaluating multiple cases (30 patients, with images from three anatomical regions per patient).
• Effect size of human readers improve with AI vs without AI assistance: This study's primary focus was on the equivalence or non-inferiority of AI-processed low-dose images compared to standard high-dose images, rather than directly measuring how much human readers improve with AI assistance over unaided interpretation of standard images. It demonstrates that AI enables a dose reduction while maintaining diagnostic performance.
  • The results show that ldDVA (AI-processed images from low-dose acquisition) reproduced ndDSA diagnostic categories with 0.84 sensitivity and 0.84 specificity.
  • Crucially, when discordant decisions were resolved by an expert, the accuracy of ndDSA and ldDVA was identical in two regions, and ldDVA outperformed ndDSA in the crural region (91% vs 80% accuracy). This implies that AI-processed low-dose images can, in some contexts, even lead to better or equivalent diagnostic outcomes by the readers compared to standard dose images, despite the dose reduction. The "improvement" is in the AI's ability to maintain or enhance quality from lower dose inputs.

6. Standalone (Algorithm Only) Performance

• Yes, standalone performance was implicitly assessed. The comparison of Contrast-to-Noise Ratio (CNR) between DSA (standard method) and DVA (Kinepict algorithm output) demonstrates an algorithmic performance advantage: "DVA produced consistently higher (two to threefold) CNR than DSA." This is a measure of the algorithm's raw image processing capability.
• Also, the improved accuracy of ldDVA in the crural region (91% vs 80%) after expert adjudication suggests the standalone algorithm's output provides superior information, even if initially some readers struggled.

7. Type of Ground Truth Used

• Expert Consensus / Expert Interpretation:
  • For the visual evaluation, the idDVA (low-dose DVA) and ndDSA (normal-dose DSA) scores from the 7 specialists served as the ground truth for comparison within that specific assessment, based on their individual expert interpretations. The comparison was for non-inferiority.
  • For the task-based diagnostic test (recognized arteries, stenosis), the "valid diagnostic category" was determined by an "expert" in cases of discordant reader decisions, implying an expert-adjudicated ground truth.
  • The "normal dose DSA images" themselves served as a de-facto ground truth (or reference standard) against which the "low-dose DVA images" were compared for equivalence in diagnostic categories.

8. Sample Size for the Training Set

• The document does not provide information regarding the sample size used for the training set of the Kinepict Medical Imaging Tool 4.0. The study described is a clinical validation study, not a development or training study.

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

• The document does not provide information regarding how the ground truth for the training set was established. This detail is typically found in design and development documentation, not necessarily in a 510(k) summary focused on clinical validation.

Ask a specific question about this device