Search Results

RayDose is software that computes dose (energy per volume deposited by ionizing radiation) three-dimensionally in a geometrical representation of a patient or a phantom, stemming from an external beam treatment unit in a radiation oncology clinic. The computed dose is intended to be used for clinical decisions on planned treatments or in quality assurance contexts.

Based on various input data, RayDose is used to compute dose for linear accelerators with X-ray energies from 6 to 18 MV, supporting the following collimation and modulation modalities:

• Symmetric and asymmetric rectangular fields
• Multileaf collimated fields
• Coplanar and non-coplanar fields
• Intensity modulated fields using Step-and-shoot technique
• Intensity modulated fields using Sliding Window technique

RayDose is a software program, which offers dose calculation, either as a separate unit or as a service to other software programs.

In short, RayDose needs the following input:

• A patient or phantom description, normally CT images and regions-of-interest
• A treatment plan
• Settings needed for the dose calculation (such as dose grid and algorithm)

RayDose computes dose to all points in the chosen dose grid, by means of applying the beams of the treatment plan onto the patient or phantom geometry. This dose computation uses the following algorithms:

• For fluence, a first principle physics based three-source fluence algorithm exists, where the first source models the collimated primary fluence, the second source models scattered and transmitted fluence from all parts of the treatment head and the third source models electron contributions to the photon fluence.
• A collapsed cone algorithm (CC) for calculating bulk doses given impinging fluence exists. This algorithm has a high accuracy also for inhomogeneities in the patient or phantom geometry.

As output, RayDose produces dose values in the chosen dose grid.

The software runs on a Windows XP platform.

1. Table of Acceptance Criteria and Reported Device Performance:

The document describes a radiation therapy dose calculation engine (RayDose). The acceptance criteria for such a device are intrinsically linked to its accuracy in calculating radiation dose. While explicit numerical acceptance criteria are not presented in a table form, the study's goal is to demonstrate that RayDose's dose calculations are comparable to or within acceptable limits of previously cleared predicate devices (DCM 1.0 and Pinnacle3 Radiation Therapy Planning System). The "reported device performance" is the assertion that RayDose performs similarly to these predicates, implying it meets their inherent accuracy standards.

2. Sample Size and Data Provenance:

The document states that the testing involved calculations for "a patient" or "phantom description" for dose calculations and "various input data" for linear accelerators. It specifically mentions using a "breast Tangent geometry (in-patient)" and a "C-shape geometry (in-patient)" test cases. Additionally, a "homogeneous phantom" was used for single field and MLC field dosimetry.

• Test Set Sample Size: The specific number of patient cases or phantom configurations used for comprehensive validation is not explicitly stated in numerical form but appears to be a small, representative set focused on specific geometries. It mentions "a patient or phantom description" and then later refers to performance for "C-shape geometry" and "breast Tangent geometry." For phantom studies, it mentions "a homogeneous phantom."
• Data Provenance: The document does not specify the country of origin for any patient data used. Given that RaySearch Laboratories AB is in Stockholm, Sweden, it is plausible that any patient data, if used, would be from Europe, though this is not confirmed. The clinical use cases mentioned (in-patient lung equivalent material) suggest that the test cases are representative of clinical scenarios. The phantom data is likely generated in a laboratory setting. The nature of the study appears to be entirely retrospective simulation and comparison based on pre-defined geometries and measurements.

3. Number of Experts and their Qualifications:

No human experts were used to establish ground truth for the test set. The validation relies on comparisons to computational reference calculations or physical measurements, not human interpretation.

4. Adjudication Method:

No adjudication method was used for the test set, as the ground truth was established through computational reference or physical measurement, not human consensus.

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

No MRMC comparative effectiveness study was done. RayDose is a dose calculation engine, not a diagnostic or decision-support tool that typically involves human readers. Its performance is evaluated against physical measurements or other calculation algorithms.

6. Standalone Performance Study:

Yes, a standalone performance study was done. The entire document describes the standalone performance of RayDose, where its dose calculations are compared to reference calculations or physical measurements without human intervention in the calculation process. The reported performance metrics (e.g., within 2% or 2mm distance-to-agreement) are direct measurements of the algorithm's accuracy.

7. Type of Ground Truth Used:

The ground truth used for the validation appears to be a combination of:

• Computational Reference: For the inpatient C-shape and breast Tangent geometries, the ground truth was established by comparison to a "reference calculation" (from a predicate device or a gold-standard calculation method, though not explicitly detailed as to which specific method).
• Physical Measurements: For single field and multileaf collimator (MLC) field dosimetry in a homogeneous phantom, the ground truth was established by "measurements." This typically refers to dosimetric measurements (e.g., using ion chambers or film) in a controlled phantom setup.

8. Sample Size for the Training Set:

The document does not provide information on a specific training set size. Radiation dose calculation engines like RayDose are typically developed using physics-based models and algorithms (e.g., collapsed cone algorithm, three-source fluence algorithm), rather than being "trained" on a large dataset of patient images in the way deep learning models are. The algorithms are derived from physical principles and validated against known physical behaviors and measurements.

9. How the Ground Truth for the Training Set was Established:

As mentioned above, RayDose's algorithms are not typically "trained" on a dataset with a conventional "ground truth" in the machine learning sense. Instead, the algorithms are developed based on established physics principles of radiation transport and interaction with matter. The parameters within these physics models are refined and validated against fundamental physical measurements and benchmarks to ensure accuracy across various conditions. Therefore, the "ground truth" for developing the underlying algorithms would be derived from:

• Physical Laws and Equations: The foundational principles of radiation physics.
• Experimental Data: Measurements from various radiation experiments to characterize beam properties, scattering, and dose deposition in different media.
• Monte Carlo Simulations: Often used as highly accurate "gold standard" computational models to verify and refine analytical dose calculation algorithms.

Ask a specific question about this device