Search Results

TRAQinform IQ is a software only device that provides a quantitative TRAQinform Report on lesions identified as Regions of Interest (ROI) in PET/CT DICOM compliant imaging data acquired, interpreted, and reported on per local practice prior to device use.

Clinicians responsible for patient care and for ordering TRAQinform Reports as an adjunct to locally reported image interpretation do not interact directly with the device. Clinicians responsible for local image interpretation do not interact with the device and generate their reporting before and independently of the TRAQinform Report.

The TRAQinform Report is generated by the device manufacturer and signed by a U.S. board certified physician responsible for supervising central report generation and qualified to practice nuclear radiology/medicine. The TRAQinform Report is for use by trained medical professionals including but not limited to oncologists, nuclear radiologists/physicians, medical imaging technologists, dosimetrists, and physicists.

TRAQinform IQ software contains the following functionalities:

• Automated matching of ROI between previously performed CT and PET/CT DICOM 3.0 volumetric medical images.
• In order to perform automated matching of ROI and quantitative analysis of previously performed CT and PET/CT DICOM 3.0 volumetric medical images, the software initially performs the following functions:
• Machine learning skeletal and anatomic structure segmentation.
• Threshold-based ROI identification and contouring.
• Automated quantitative analysis to assess previously performed CT and PET/CT DICOM 3.0 volumetric medical images, including: change in total volume and density of each identified ROI, and change in Fludeoxyglucose F18 (FDG) tracer uptake of each identified ROI among images.
• Generation of images of the anatomy combined with spatial and quantitative information, including computed classification of quantitative FDG ROI changes.

For multi-timepoint quantitative analysis, recommended use is in adult patients 22 years and older with partial or whole-body PET/CT acquired following administration of FDG per approved drug prescribing information and with the second FDG administration separated from the first by a period not to exceed 12 months.

For single-timepoint quantitative analysis, recommended use is in adult patients 22 years and older with partial or whole-body PET/CT following administration of FDG, a PSMA targeted PET drug, or a SSTR-targeted PET drug per approved drug prescribing information.

Discrepancies between TRAQinform IQ and local PET/CT reporting have been investigated and use of TRAQinform IQ has not been established for binary patient level progression or non-progression decisions without multidisciplinary review. Discrepancies between TRAQinform IQ and local PET/CT reporting that could impact patient care should therefore prompt consultation with subject matter experts (for example, in tumor board), with a patientcentered focus on discrepant imaging regions and with blinded or otherwise neutral adjudication regarding interpretation/classification source.

TRAQinform IQ is not intended to diagnose any disease, replace the diagnostic procedures for interpretation of CT or PET/CT images, recommend any specific treatment, nor is it intended to replace the skill and judgment of a qualified medical professional.

TRAQinform IQ is a software only device that provides quantitative analysis of lesions identified as Regions of Interest (ROI) in PET/CT DICOM compliant imaging data acquired, interpreted, and reported on per local practice prior to device use.

The input to TRAQinform IQ is CT and PET/CT images as supported by ACR/NEMA DICOM 3.0.

The following steps are performed by the software:

• Automatic threshold-based ROI segmentation:
• ROI can also be imported from external sources (other validated tools or manual contouring by qualified medical personnel).
• Automatic ROI registration between multiple images:
• Images can be from the same or different imaging modality.
• Images can be from the same or different PET tracer.
• Images can be from the same or different date.
• Automatic matching of ROI between multiple, previously performed images.
• Automatic quantification of dynamic changes among images including, but not limited to:
• Changes in ROI shape.
• Single ROI splitting into multiple ROI.
• Multiple ROI combining into a single ROI.
• ROI appearing, disappearing, and re-appearing across images.
• A comprehensive summary analysis.

TRAQinform IQ calculates spatial and quantitative metrics for each individual ROI. These metrics are provided as a TRAQinform Report. TRAQinform IQ uses computational algorithms to detect, fuse and analyze ROI and provides the following outputs:

• Identification of anatomic location of ROI in all areas of the body.
• A quantitative analysis of functional and anatomic data for CT and PET/CT scans, including:
• Volume of all identified ROI on each image;
• Change in volume of each identified ROI among images:
• Total volume of all identified ROI on each image;
• Change in total volume of all identified ROI on each image;
• Heterogeneity of change in volume of each identified ROI:
• For PET scans:
• Tracer uptake (SUVmax, SUVtotal, SUVmean, SUVhetero) of each identified ROI on each image:
• Change in tracer uptake (SUVmax, SUVtotal, SUVmean) of each identified ROI among images:
• Total tracer uptake (SUVmax, SUVtotal, SUVmean, SUVhetero) of all identified ROI on each image:
• Change in total tracer uptake (SUVmax, SUVtotal, SUVmean, SUVhetero) of all identified ROI on each image:
• Heterogeneity of change in tracer uptake (SUVhetero) among identified ROI.
• For CT scans
• Radio density (HUmax, HUtotal, HUmean, HUhetero) for each identified ROI on each image:
• Changes in radio density (HUmax. HUtotal. HUmean) of all identified ROI on each image:
• Change in total radio density (HUmax, HUtotal, HUmean, HUhetero) of all identified ROI on each image;
• Heterogeneity of change in radio density (HUhetero) among identified ROI.
• 2D graphical renderings of medical images, including Maximum Intensity Projections of the PET and CT, with overlayed and labeled/color-coded ROI, for inclusion in TRAQinform Reports.
• 3D labeled contours for ROI, anatomic structures, and skeletal structures.

The TRAQinform IQ software operates in a secure cloud environment.

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

Acceptance Criteria and Device Performance Study for TRAQinform IQ

1. Acceptance Criteria and Reported Device Performance

The provided text summarizes performance data from two studies: a "Test-Retest" reliability study and a "Pivotal Reader Study." The acceptance criteria, while not explicitly stated as "acceptance criteria" for regulatory submission, can be inferred from the reported performance measures and context of the studies.

Test-Retest Study: Limits of Repeatability for Quantitative Features

This study established the expected variability of the device's quantitative measurements. The limits of repeatability serve as an informal "acceptance criteria" for the intrinsic variability of the measurements.

Reported Device Performance: The table above is the reported device performance for the test-retest study, indicating the interval within which 95% of repeated measurements are expected to lie.

Pivotal Reader Study: Agreement with Expert Panel

This study evaluated the clinical utility of TRAQinform IQ by assessing how well its output, when presented to oncologists, aligned with an expert panel's assessment. The "acceptance criteria" here would be an adequate level of agreement, though specific thresholds are not explicitly defined as pass/fail.

Note on Acceptance: The document states that the "performance data demonstrate that the TRAQinform IQ is as safe and effective as the QTxl." This implies that the reported performance metrics were deemed acceptable for substantial equivalence.

2. Sample Sizes and Data Provenance

Test-Retest Study:

• Sample Size: 31 patients.
• Data Provenance: Patients with non-small cell lung cancer, received two FDG PET/CT images within 1 week pretreatment. No explicit location (e.g., country) is given, but the context of an FDA submission suggests data highly relevant to the US market. The study design (two scans within 1 week) indicates a prospective, controlled data collection for evaluating reliability.

Pivotal Reader Study:

• Sample Size: 103 patients, each with two sequential FDG PET/CT scans (total 206 scans).
• Data Provenance: Images acquired between 2005 and 2022 from patients scanned at 10 or more imaging centers in at least 3 U.S. states. This indicates retrospective data collection from real-world clinical practice in the USA. Specific scanner information (manufacturers and models) is provided, and 84 patients had scans on the same scanner for baseline and follow-up. Patient demographics (cancer type, sex, age, weight, race) are also detailed.

3. Number of Experts and Qualifications for Ground Truth

Test-Retest Study:

• The ground truth for this study was based on the device's own measurements. There is no mention of expert image interpretation being used to establish a ground truth for "limits of repeatability."

Pivotal Reader Study:

• Number of Experts: A panel of three experts was used.
• Qualifications of Experts: Two radiologists and one oncologist. No specific experience levels (e.g., "10 years of experience") are explicitly given, but their titles (radiologist, oncologist) imply qualified medical professionals in their respective specialties.

4. Adjudication Method for the Test Set

Pivotal Reader Study (for expert panel ground truth):

• The data states: "Imaging and local reporting on these 23 + 9 = 32 patients was sent to a panel of two radiologists and one oncologist, together serving as a reference source against which to quantify..." This suggests a consensus-based adjudication method (all three experts together formed the reference source), rather than a majority rule or other multi-reader approach. The document doesn't specify if it was a "2+1" or "3+1" approach, but it implies a collective decision by the panel.

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

• Yes, a form of MRMC study was done. The "Pivotal Reader Study" involved three oncologist "report evaluators" reading cases without and then with the adjunctive TRAQinform Report.

• Effect Size of Human Reader Improvement: The study demonstrates how the AI assistance changes human reader interpretations.

  • 23 patients initially classified as "negative for progression" by oncologists without the device were reevaluated to "positive" with the device.
  • 9 patients initially classified as "positive for progression" by oncologists without the device were reevaluated to "negative" with the device.

  This indicates that the AI report prompted a re-evaluation and change in classification for 32 out of 103 patients (approx. 31%). The "effect size" is the shift in clinical decision, influencing a significant percentage of cases. The PPA and NPA against the expert panel further quantify the agreement (or disagreement) after AI assistance. The key finding is the change in oncologist assessment, even if not directly framed as an "improvement" in accuracy by the text itself, but rather as influencing the decision toward what the expert panel considered ground truth.

6. Standalone (Algorithm Only) Performance

• Yes, standalone performance aspects were evaluated indirectly.
  • The "Test-Retest" study primarily assesses the standalone stability and reproducibility of the algorithm's quantitative measurements (SUVmax, SUVmean, SUVtotal, Volume) without human interaction influencing the values themselves.
  • The "agreement... with the device classification" for ROIs highlighted by the report evaluators (70-90% for various ROI changes) is also a measure of the algorithm's performance against the expert panel's assessment. This isn't a full "standalone diagnostic accuracy" but rather an evaluation of the algorithm's output (classification of ROI changes) compared to the panel.
  • The text also references "early feasibility testing of device component functionality," with published reporting available for CT anatomy segmentation (Weisman 2023), ROI detection methodology (Perk 2018), and ROI matching (Huff 2023). While not detailed here, these studies would likely have included standalone performance evaluation for those specific algorithmic components.

7. Type of Ground Truth Used

• Expert Consensus: For the "Pivotal Reader Study," the primary ground truth for patient-level progression assessment (PPA, NPA, OPA) was established by a panel of two radiologists and one oncologist acting as a "reference source."
• Inherent Ground Truth (device's own outputs): For the "Test-Retest" study, the ground truth for repeatability was the device's own measurements across repeated scans, assessed for consistency.

8. Sample Size for the Training Set

• The document does not explicitly state the sample size for the training set used for the TRAQinform IQ algorithm.
• It does mention that the software uses "Machine learning skeletal and anatomic structure segmentation" (page 4, also page 7, 8). While the exact training dataset size isn't listed, the reference to published papers (Weisman 2023, Perk 2018, Huff 2023) suggests that the underlying machine learning components would have been trained using relevant datasets as described in those publications.

9. How Ground Truth for Training Set was Established

• The document does not explicitly describe how the ground truth for the training set was established.
• Given the mention of "Machine learning skeletal and anatomic structure segmentation" and "Threshold-based ROI identification and contouring," the ground truth for training these models would typically involve expert annotations of anatomical structures and ROIs on medical images. This would be consistent with standard practices for training medical image segmentation and detection algorithms. The referenced external publications (Weisman 2023, Perk 2018, Huff 2023) would contain details on their specific training methodologies and ground truth establishment.

Ask a specific question about this device