(143 days)
Hybrid3D that provides software applications used to process, display, and manage nuclear medicine and other medical imaging data transferred from other workstation or acquisition stations.
HERMES Hybrid3D is a reading and processing module for the advanced needs in medical imaging. It offers multi-modal (PET/CT/MR/SPECT) coregistration and interactive fusion of multiple datasets. HybridViewer 3D handles viewing and fusion of multi-sequence MRI studies with oblique orientation and allows switching between original and standard TCS view orientation as well as defining own slice directions. 3D seqmentation, cropping and interpolation techniques allow complex tasks in VOI definition and can cover cases like cavities, splitting structures into subsections or logic operations (compute intersections, merge, grow). Results can be imported and exported as DICOM and are therefore available for research in 3rd party tools. Additionally, it provides tools for advanced 3D fusion rendering of studies and VOIs.
Lung Lobe Quantification: The Lung Lobe Quantification module in Hybrid3D, introduces an efficient and automated workflow solution to accurately compute 3D lobar anatomy from CT (with or without contrast). The workflow supports the addition of functional images (SPECT V/Q, SUV SPECT, CT iodine maps, hyperpolarized xenon MRI, etc.) to accurately relate lobar anatomy to function. No changes have been made to Lung Lobar Quantification since the previous release.
TumorFinder: The Tumor Finder wizard provides automatic segmentation of lesions in a PET study or a combined PET/CT study pair, based on criteria relative to a background volume placed in the liver or mediastinum. This reduces the time required for tumor delineation. It also provides both visual and statistical evaluation of tumor burden, which helps with comparing follow up studies.
SIRT: Selective Internal Radionuclide Therapy (SIRT), is currently used in the treatment of liver tumors either from primary liver cancer or metastatic disease (e.g. colorectal primary cancer). The SIRT wizard provides processing for SIRT planning and verification.
The provided text is a 510(k) summary for the medical device Hybrid3D v3.0. While it discusses software features, regulatory details, and some testing, it does not contain a detailed study proving the device meets specific acceptance criteria in the manner typically expected for AI/Machine Learning-based medical devices.
Instead, the performance evaluation in this document focuses on:
- Comparison to a predicate device (Hybrid3D v2.0): Stating "The proposed device will use similar technology and fundamental concepts and operation are also the same," and "The comparisons between Hybrid 3D v3.0 and Hybrid 3D v2.0 (K171719) were part of the test procedure for V3.0 and showed good results." This implies a functional equivalence rather than a new clinical performance study.
- Validation of specific calculations for the SIRT module: This is a validation of the accuracy of mathematical computations within a specific module, not a broad clinical performance assessment of features like image processing or tumor finding from AI.
Therefore, for many of your specific questions, the information is not present in the provided text. I will address what is available and clearly state what is missing.
Here's a breakdown based on the provided text:
1. Acceptance Criteria and Reported Device Performance
The document does not present a formal table of general acceptance criteria for the entire Hybrid3D device, nor does it provide a comprehensive "reported device performance" in terms of clinical metrics (e.g., sensitivity, specificity, accuracy).
However, it does provide implicit acceptance criteria and reported "performance" for the SIRT (Selective Internal Radionuclide Therapy) module's calculations:
| Acceptance Criteria (Implicit) | Reported Device Performance (SIRT Module Calculations) |
|---|---|
| Lung Shunt calculations accuracy | Identical to spreadsheet calculations. |
| Prescribed Activity (Resin Microspheres, BSA method) | Identical to 2 decimal places to spreadsheet calculations. |
| Activity to Implant (Glass Microspheres, PTV method) | Identical to 2 decimal places to spreadsheet calculations. |
| Voxel Dose (PETDose Map) accuracy | Identical to 2 decimal places to spreadsheet calculations. |
| Voxel Dose (SPECT Dose Map) accuracy | Varied by up to 5% compared to spreadsheet calculations (due to normalization differences). |
| Absence of fundamental errors in Vmax and D90% calculations (based on manual reading challenges) | Established that there were no fundamental errors in the calculations, despite noted variances due to manual reading inaccuracy. |
2. Sample Size and Data Provenance
The document does not specify sample sizes (e.g., number of patients, number of images) used for any test set or the provenance (country of origin, retrospective/prospective) of any data used for testing. The validation described for the SIRT module appears to be a comparison of calculation results against a spreadsheet, not a clinical data set.
3. Number and Qualifications of Experts for Ground Truth
The document does not mention the use of experts or their qualifications for establishing ground truth for any test set, as the described validation is for calculation accuracy against spreadsheet formulae, not expert interpretation of images.
4. Adjudication Method
The document does not mention any adjudication method, as it does not describe a process involving multiple readers or complex ground truth establishment.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
The document does not indicate that an MRMC comparative effectiveness study was performed, or any effect size of AI assistance on human readers. The device is primarily described as a software for processing, display, and management of imaging data, and specific quantification modules, not an AI-assisted diagnostic aid that directly impacts human reader performance in a comparative study.
6. Standalone (Algorithm Only) Performance
The document describes the device's functional capabilities (image processing, display, quantification for SIRT, TumorFinder, Lung Lobe Quantification), but it does not present a standalone performance study (e.g., sensitivity, specificity, accuracy) for these algorithmic features, especially for TumorFinder or Lung Lobe Quantification, against a clinical ground truth. The "testing results supports that all the software specifications have met the acceptance criteria" is a very general statement. The specific validation described is for the calculation accuracy of the SIRT module.
7. Type of Ground Truth Used
For the specific validation described for the SIRT module, the "ground truth" used was:
- Spreadsheet calculations based on formulae published by SIRTEX for Resin Microspheres and BTG for Glass Theraspheres. This is a form of scientific/mathematical ground truth for the accuracy of internal calculations, not a clinical ground truth like expert consensus, pathology, or outcomes data.
For other modules like TumorFinder or Lung Lobe Quantification, the document does not describe how their performance was validated or what type of ground truth was used.
8. Sample Size for Training Set
The document does not mention any training set sample size, which suggests that the development did not involve a machine learning model that required a distinct training phase in the context of this 510(k) submission. Given the description focusing on image processing, co-registration, 3D segmentation, and rule-based quantification (TumorFinder "based on criteria relative to a background volume"), it's plausible the "AI" aspects are more algorithmic and rule-based rather than deep learning requiring large training sets.
9. How Ground Truth for Training Set Was Established
Since no training set is mentioned, this information is not applicable/provided.
§ 892.1200 Emission computed tomography system.
(a)
Identification. An emission computed tomography system is a device intended to detect the location and distribution of gamma ray- and positron-emitting radionuclides in the body and produce cross-sectional images through computer reconstruction of the data. This generic type of device may include signal analysis and display equipment, patient and equipment supports, radionuclide anatomical markers, component parts, and accessories.(b)
Classification. Class II.