Search Results

LiverMultiScan v5 (LMSv5) is indicated for use as a magnetic resonance diagnostic device software application for noninvasive liver evaluation that enables the generation, display and review of 2D magnetic resonance medical image data and pixel maps for MR relaxation times.

LMSv5 is designed to utilize DICOM 3.0 compliant magnetic resonance image datasets, acquired from compatible MR Systems, to display the internal structure of the abdomen including the liver. Other physical parameters derived from the images may also be produced.

LMSv5 provides several tools, such as automated liver segmentation and region of interest (ROI) placements, to be used for the assessment of selected regions of an image. Quantitative assessment of selected regions includes the determination of triglyceride fat fraction in the liver (PDFF), T2*, LIC (Liver Iron Concentration) and iron corrected T1 (cT1) measurements.

These images and the physical parameters derived from the images, when interpreted by a trained clinician, yield information that may assist in diagnosis.

LiverMultiScan v5 (LMSv5) is a standalone software device. The LiverMultiScan v5 device is to assist a trained operator with the evaluation of information from Magnetic Resonance (MR) images from a single time-point (a patient visit).

LiverMultiScan is a post-processing software device, a trained operator uses tools within the device interface to quantify liver tissue characteristics from parametric maps. LiverMultiScan v5 includes automatic processing functionality based on machine-learning to assist in the quantification of metrics during analysis, such as automatic artefact detection and automatic segmentation of the liver. A summary report from the analysis conducted is generated for interpretation by a clinician.

LiverMultiScan v5 is not intended to replace the established procedures for the assessment of a patient's liver health by a clinician, providing many opportunities for competent human intervention in the clinical care of patients.

The metrics are intended to be used as an additional diagnostic input to provide information to clinicians as part of a wider diagnostic process. It is expected that in the normal course of clinical care, patients with clinical symptoms or risk factors which may indicate liver disease. The interpreting clinician needs to take into consideration the device's limitations and accuracy during clinical interpretation.

Information gathered through existing diagnostic tests and clinical evaluation of the patient, as well as information obtained from LiverMultiScan v5 metrics, may contribute to a diagnostic decision.

LiverMultiScan v5 is not a computer-aided diagnostic device and can only present imaging information which must be interpreted by a qualified clinician. LiverMultiScan v5 is an aid to diagnosis and treatment decisions remain the responsibility of the clinician.

Here's an analysis of the acceptance criteria and study information for LiverMultiScan v5 (LMSv5) based on the provided text, structured according to your requested points:

1. Table of Acceptance Criteria and Reported Device Performance

The provided FDA 510(k) summary document does not explicitly list quantitative acceptance criteria in a dedicated table format. Instead, it states that "All product specifications were verified and the overall ability of the product to meet user needs was validated." and "The accuracy and precision of device measurements was assessed using purpose-built phantoms...". It also mentions "The performance testing conducted demonstrates that LiverMultiScan v5 is at least as safe and effective as the predicate devices."

While specific numerical acceptance criteria (e.g., "PDFF accuracy must be within X%", "LIC precision must be within Y%") are not detailed in this public document, the general acceptance can be inferred from the statement that the device meets safety and effectiveness standards comparable to its predicates. The performance claims are therefore that the device provides accurate and precise measurements of liver triglyceride fat fraction (PDFF), T2*, Liver Iron Concentration (LIC), and iron-corrected T1 (cT1), comparable to or exceeding its predicate devices.

2. Sample size used for the test set and the data provenance

• Test Set Sample Size: The document mentions "in-vivo volunteer data was used" for assessing precision across supported scanners and for inter/intra-operator variability. It also refers to "purpose-built phantoms" for accuracy and precision of device measurements. Specific numerical sample sizes for either phantom or in-vivo data are not provided in this summary document.
• Data Provenance: Not explicitly stated from which country the "in-vivo volunteer data" originated. It is generally understood that studies supporting medical device submissions to the FDA can involve multinational data, but this document does not specify. The data would be prospective if volunteers were recruited and scanned specifically for this validation, or retrospective if existing de-identified patient data was used. The document does not clarify this.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

This document does not provide details on the number or qualifications of experts used to establish ground truth for the test set. For quantitative MRI biomarkers like PDFF, T2*, cT1, and LIC, ground truth is typically established through direct physical measurements (e.g., chemical extraction for fat, histology/biopsy for iron/fibrosis, though biopsy has limitations) or well-validated reference standards (phantoms or highly accurate established MRI methods). The document mentions "purpose-built phantoms containing vials with different relaxation times corresponding to the physiological ranges of tissue values" were used for accuracy and precision, implying these phantoms served as a form of ground truth for those measurements.

4. Adjudication method for the test set

The document does not describe any expert adjudication methods (e.g., 2+1, 3+1) for the test set within this summary. For this type of quantitative imaging device, the focus is more on technical accuracy and precision against established physical or biological references rather than subjective expert consensus on image interpretation.

5. If a multi-reader multi-case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

The document does not report an MRMC comparative effectiveness study where human readers' performance with and without AI assistance (LMSv5) was evaluated. The device is described as "post-processing software" that provides "additional diagnostic input" and is "an aid to diagnosis." It includes "automatic processing functionality based on machine-learning to assist in the quantification of metrics... such as automatic artefact detection and automatic segmentation of the liver." However, the study described focuses on the device's standalone performance (accuracy, precision, variability) and not on its impact on human reader performance in a controlled comparative setting.

6. If a standalone (i.e., algorithm only without human-in-the-loop performance) was done

Yes, a standalone performance assessment was done. The testing described focuses on the device's intrinsic capabilities:

• "The accuracy and precision of device measurements was assessed using purpose-built phantoms..."
• "To assess the precision of LiverMultiScan v5 measurements across supported scanners, in-vivo volunteer data was used."
• "Inter and intra operator variability was also assessed." (While this involves human operators, it assesses the device's consistency under different human interactions, not the human's diagnostic performance).
• The device is a "standalone software device" and "All operations are directly controlled by the LiverMultiScan device."

This indicates that the technical performance of the algorithm itself, in generating the quantitative metrics (PDFF, T2*, LIC, cT1), was evaluated.

7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.)

The primary ground truth described in this summary are:

• Physical Phantoms: "purpose-built phantoms containing vials with different relaxation times corresponding to the physiological ranges of tissue values expected to be seen in-vivo" were used to assess the accuracy and precision of measurements. This provides a quantifiable, objective ground truth for the relaxation times and component fractions (like fat).
• For LIC, the device "uses the measured T2* value and uses them to characterise iron loading in the liver which is then transformed by a defined calibration curve to provide a quantitative measure of liver iron concentration in vivo." This implies the ground truth for establishing that calibration curve would likely have been against biopsy-derived liver iron concentration data or other highly validated reference methods in earlier studies (though not explicitly stated for this particular submission's testing itself, it's the basis for LIC quantification).

8. The sample size for the training set

The document does not specify the sample size used for the training set. It mentions that LMSv5 includes "automatic processing functionality based on machine-learning to assist in the quantification of metrics during analysis, such as automatic artefact detection and automatic segmentation of the liver." This clearly indicates machine learning was used, and thus a training set was necessary. However, the size and characteristics of that training set are not included in this 510(k) summary.

9. How the ground truth for the training set was established

The document does not detail how the ground truth for the training set (used for machine learning components like automated segmentation and artifact detection) was established. Typically, for such features, ground truth would be established through:

• Expert manual segmentation/annotation: Radiologists or trained medical image analysts would manually delineate liver boundaries and artifacts on a large dataset of MR images.
• Consensus from multiple experts: To reduce individual bias, several experts might review and agree upon segmentations or artifact labeling.

However, these specifics are omitted from this summary.

Ask a specific question about this device