(83 days)
Velocity is a software package that provides the physicians a means for comparison of medical data including imaging data that is DICOM compliant. It allows the display, annotation, volume rendering, registration, and fusion of medical images as an aid during use by diagnostic radiology, radiation therapy planning and other medical specialties. Velocity is not intended for mammography.
Velocity is a software application providing relevant tools for Radiotherapy professionals to use while creating treatment plans. The Velocity device is a Picture Archiving and Communication System (medical imaging software). Velocity provides medical image processing designed to facilitate the oncology or other clinical specialty work flow by allowing the comparison of medical imaging data from different modalities, points in time, and / or scanning protocols. The product provides users with the means to display, co-register and analyze medical images from multiple modalities including PET, SPECT, CT, RT Dose and MR; draw Regions of Interest (ROI), calculate and report relative differences in pixel intensities, Standardized Uptake Value (SUV) or other values within those regions; and import / export results to/from commercially available radiation treatment planning systems and PACS devices. Co-registration includes deformable registration technology that can be applied to DICOM data including diagnostic and planning image volumes, structures, dose, and automatic anatomical atlas-based segmentation tools. Velocity is used as a stand-alone application on recommended Off-The-Shelf (OTS) computers supplied by the company or by the end-user.
The provided document is a 510(k) Premarket Notification for the "Velocity" software. It focuses on demonstrating substantial equivalence to a predicate device ("Velocity AI – K081076") and does not contain detailed information about specific acceptance criteria, test set sizes, expert qualifications, or comparative effectiveness studies typically associated with proving a device meets strict performance benchmarks via a clinical study with AI components.
However, based on the information provided, we can infer some aspects and highlight what is explicitly stated:
Key Takeaways from the Document:
- Device Type: Velocity is a "Picture Archiving and Communication System (Medical Imaging Software)" designed for display, annotation, volume rendering, registration, and fusion of medical images for diagnostic radiology, radiation therapy planning, and other medical specialties.
- Predicate Device Approach: The submission relies on demonstrating "substantial equivalence" to a legally marketed predicate device (Velocity AI – K081076) and a secondary predicate for a specific feature (MIM Y90 Dosimetry - K172218).
- No Clinical Tests: The document explicitly states: "No clinical tests have been included in this pre-market submission." This immediately tells us that the detailed performance studies often associated with AI/ML model validation (e.g., MRMC studies, standalone performance with robust ground truth) were not part of this submission for the core functionality.
- Software Verification & Validation (V&V): The primary performance data cited is "Software Verification and Validation Testing," indicating a focus on functional correctness, reliability, and safety of the software rather than a direct clinical performance evaluation. The software was considered a "major" level of concern.
- Ground Truth: Given no clinical tests, there's no mention of how ground truth for a test set was established using expert consensus, pathology, or outcomes data for clinical performance metrics. The ground truth for V&V would be tied to software requirements and specifications.
- AI/ML Context: While "AI" is in the predicate device name ("Velocity AI"), the listed functionalities of "Velocity" (display, annotation, volume rendering, registration, fusion) are standard PACS and image processing features. The "deformable registration technology" and "automatic anatomical atlas-based segmentation tools" could be considered AI/ML elements, but the document does not detail specific performance studies of these elements with clinical ground truth.
Based on the available text, here's an attempt to answer your questions. Please note that many answers will state that the information is "Not provided" or "Not applicable" due to the nature of this 510(k) submission, which did not include clinical performance studies.
1. A table of acceptance criteria and the reported device performance
The document does not provide a specific table of acceptance criteria with corresponding reported device performance metrics in the way a clinical study typically would (e.g., sensitivity, specificity, F1-score for an AI component). The performance data cited is:
| Acceptance Criteria Category | Nature of Performance Demonstrated | Reported Device Performance |
|---|---|---|
| Functional Equivalence | Comparison to predicate device | "Velocity 4.0 performs similar to the predicate device VelocityAIS v2.0 (K081076) for the functions contained within the predicate" (Implied from comparison table and substantial equivalence claim). |
| Safety & Effectiveness | Software Verification & Validation | "The non-clinical data support the safety of the device and the software verification and validation demonstrate that the Velocity device performs as intended." |
| Specific Feature Equivalence (Y90 Dosimetry) | Comparison to secondary predicate | "MIM Y90 Dosimetry (K172218) was determined to be a predicate for the RapidSphere SPECT Microsphere Dosimetry feature. Both features: are intended for post-treatment absorbed dose calculation and evaluation, are compatible with Y90 PET and SPECT image types, and use the Local deposition model for dose calculation." |
2. Sample sized used for the test set and the data provenance (e.g. country of origin of the data, retrospective or prospective)
- Sample Size for Test Set: Not specified. As "No clinical tests have been included," there isn't a "test set" in the context of clinical performance evaluation using patient data for AI validation. The V&V would use internal test cases.
- Data Provenance: Not specified.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts (e.g. radiologist with 10 years of experience)
- Not applicable. Since no clinical tests were performed for the submission, there is no mention of external experts establishing ground truth for a clinical test set.
4. Adjudication method (e.g. 2+1, 3+1, none) for the test set
- Not applicable. No clinical tests were performed.
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
- No. The document explicitly states: "No clinical tests have been included in this pre-market submission." Therefore, an MRMC comparative effectiveness study was not performed or submitted.
6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done
- Not specified in a clinical performance context. While the software operates "as a stand-alone application," this refers to its deployment rather than a standalone algorithmic performance study against clinical ground truth. The V&V process would have involved testing the algorithm's functional performance against specified requirements.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)
- For Software Verification and Validation: The ground truth would be defined by the software requirements and specifications (e.g., expected output for given input, correctness of calculations, proper display of images). This is a technical ground truth, not a clinical one derived from patient outcomes or expert consensus on a diagnosis.
- For Clinical Performance: Not applicable, as no clinical tests were performed.
8. The sample size for the training set
- Not specified. The document does not describe the development or training of any AI/ML components beyond mentioning "deformable registration technology" and "automatic anatomical atlas-based segmentation tools." No information is provided regarding the training data or its size.
9. How the ground truth for the training set was established
- Not specified. As there is no information on a specific training set or the development process for embedded AI/ML features, how any training ground truth was established is not detailed.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).