Search Results

Torch™ software is intended to provide estimates (deterministic) of absorbed radiation dose at the voxel level following internal administration of approved radioactive products. This is dependent on input data regarding biodistribution being supplied to the application. Torch software only allows voxel-based dose calculations. For use with internally administered radioactive products. Torch should not be used to deviate from approved product dosing and administration instructions. Refer to the product's prescribing information for instructions.

Torch is a software device for use in absorbed dose estimation of FDA approved radiopharmaceuticals. After administration of FDA approved radiopharmaceuticals, Torch provides post-treatment dosimetry evaluation on the voxel level using Monte Carlo techniques to assist the clinician in assessing absorbed dose to normal tissues and tumors.

Key functions of the Torch software include image import, image registration, contour propagation, pharmacokinetic modelling, and radiation transport modelling which is accelerated using parallel processing capabilities of Graphics Processing Units (GPUs). A Monte Carlo (MC) radiation transport algorithm is used to calculate absorbed dose with very high dosimetric accuracy.

Torch relies on medical images (e.g., positron emission tomography (PET), single photon emission computed tomography (SPECT), or computed tomography (CT)) which consist of a three-dimensional matrix of pixels, called voxels. Medical imaging acquired on the voxel level affords measurement of tissue heterogeneity and nonuniform source distribution of radiopharmaceuticals. Monte Carlo produces dose distributions at the voxel level with high precision.

Here's an analysis of the acceptance criteria and the study that proves the device meets them, based on the provided text:

1. Table of Acceptance Criteria and Reported Device Performance

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

• Sample Size for Test Set: The "ICRP 110 Adult Male reference dataset" was used. The number of specific organs or dose calculations performed within this dataset is not explicitly quantified as a "sample size" in the traditional sense of a patient cohort. It refers to a standardized computational phantom.
• Data Provenance: The ICRP 110 Adult Male reference dataset is a widely recognized computational model, not data derived from retrospective or prospective human participant studies. The physical film measurements were conducted in collaboration with the University of Wisconsin Accredited Dosimetry Calibration Laboratory, suggesting laboratory-generated data.

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

• The ground truth for the primary tests was established by comparison with three established Monte Carlo codes: EGS++ 2018, GATE 8.1, and GEANT4 10.5 (i.e. the OpenDose project). These are not "experts" in the human sense, but rather established and validated computational models that serve as a reference standard.
• For the physical film measurements, "physical film measurements" were used, implying a measured ground truth, likely validated by experts in dosimetry at the University of Wisconsin. The specific number and qualifications of experts involved in the film measurements are not provided.

4. Adjudication Method for the Test Set

• The adjudication method involved direct comparison of Torch's results against the average results from the three established Monte Carlo codes (OpenDose project) and against OLINDA version 2.0. For the film measurements, it was a direct comparison against the measured data and an EGS algorithm. There isn't an explicit "adjudication" in the sense of multiple human annotators; rather, it's a quantitative comparison to pre-defined reference standards or measurements.

5. If a Multi Reader Multi Case (MRMC) Comparative Effectiveness Study Was Done

• No, a Multi Reader Multi Case (MRMC) comparative effectiveness study was not done. The document explicitly states: "No clinical studies were required for validation of the Torch software." The evaluation focuses on the algorithm's accuracy against established computational models and physical measurements, not on human reader performance with or without AI assistance.

6. If a Standalone Study Was Done

• Yes, a standalone (algorithm only without human-in-the-loop performance) study was done. The entire performance data section describes the evaluation of the "Torch absorbed dose" comparing it directly to reference codes and measurements. The device is intended to provide estimates of absorbed dose, and the testing validates the accuracy of these estimations directly.

7. Type of Ground Truth Used

• The ground truth used was a combination of:
  • Expert Consensus/Established Computational Models: The average results from the three established Monte Carlo codes (EGS++ 2018, GATE 8.1, GEANT4 10.5, i.e., the OpenDose project) and OLINDA version 2.0. These codes represent widely accepted and validated methods in dosimetry.
  • Physical Measurements: For one set of tests, physical film measurements were used as ground truth.

8. Sample Size for the Training Set

• The document does not explicitly mention a training set sample size. The nature of the validation focuses on comparison to established physical and computational models, suggesting a validation of the underlying physical model and implementation, rather than a machine learning model that requires a distinct training and test set with ground truth labels. If the "Monte Carlo techniques" involve any machine learning components, details about their training are not provided.

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

• Since a training set is not explicitly mentioned or described as part of the validation, the method for establishing its ground truth is not provided in the document.

Ask a specific question about this device