(191 days)
RemedyLogic AI MRI Lumbar Spine Reader ("RAI") is an image post-processing and measurement software tool that provides quantitative spine measurements from previously-acquired DICOM lumbar spine Magnetic Resonance (MR) images for users' review, and interpretation. It provides the following functionality to assist users in visualizing, measuring and documenting measurements:
· Feature segmentation;
· Feature measurement; and
· Exportation of measurement results in DICOM Structured Report and a downloadable .docx file for users to review and to use full or partial list of software-generated measurements to prepare their own radiology report.
RAI does not produce or recommend any type of medical diagnosis or treatment. Instead, it simply helps users to more easily identify and measure features in lumbar MR images and compile their own reports. The user is responsible for reviewing, verifying, and correcting, if necessary, the software-generated segmentations and measurements, leveraging useful software output and using their medical judgment and discretion to make diagnostic or treatment decisions.
The device is intended to be used only by radiologists, neuro- and spine-surgeons in hospitals and other medical institutions.
Only T2 MRI images in DICOM format, acquired from lumbar spine exams of patients aged 18 and older, are considered to be valid input. RAI does not support DICOM images of patients that are pregnant, undergo MRI scan with contrast media, or have post-operational complications, scoliosis, tumors, infections, and/or fractures.
The RemedyLogic AI MRI Lumbar Spine Reader (RAI) is an MR image post-processing and measurement software tool that provides quantitative spine measurements from previously acquired DICOM lumbar spine Magnetic Resonance (MR) images for qualified users' review, analysis, and interpretation. The qualified users (i.e., radiologists, spine- and neuro-surgeons) are physicians qualified to read and interpret spine MRI exams in a manner consistent with American College of Radiology (ACR) recommendations.
The RAI analyzes the user-uploaded lumbar spine images and provides the following functionalities to assist qualified users in visualizing images, and measuring images, and generating reports:
- Feature segmentation: the software automatically detects the borders of anatomical objects of interest and generates the corresponding contours for these objects.
- 0 Feature measurement: the software automatically generates common measurements for segmented objects.
- 0 Measurement export: a DICOM Structured Report or a .docx file containing the measurement results can be exported for users to review and to use the full or a partial list of the softwaregenerated measurements to prepare their own radiology reports.
The RAI software does not interface directly with any MR scanner or data collection equipment. Rather, a qualified user must upload a previously acquired MR study in DICOM format into the RAI software via their Picture Archiving and Communication System (PACS). The PACS serves as the RAI user interface. After less than two minutes of processing, the RAI software automatically generates and uploads back to PACS the DICOM with segmentations of regions of interest along with corresponding measurements. These measurements are also presented in a DICOM Structured Report and a downloadable .docx file, which is accessible for download from the PACS from PACS system. The user reviews the softwaregenerated measurement list, studies the software-annotated images and/or the original unannotated images when necessary, and reviews other pertinent medical information about the patient. The user can manually segment anatomical objects and mark their own measurements using the DICOM viewer tools. The user can also edit measurements in the downloaded .docx file. The user then writes their own radiology report, incorporating some or all verified or corrected measurements, with diagnosis and/or treatment recommendations.
The purpose of the RAI software is to save time by automating tedious, time-consuming, and potentially error-prone manual tasks. The software does not perform any functions that could not be accomplished by a qualified user. The outputs of the software, i.e. feature segmentations and quantitative measurements, are reviewed, analyzed, confirmed or corrected by the user before any such content is included in the user's final report.
The RAI consists of a cloud-based machine learning (ML) analytical algorithm deployed on a GPU cloud service and an API to integrate directly with the client's PACS system.
Here's a summary of the acceptance criteria and the study proving the device meets them, based on the provided text:
Device Name: RemedyLogic AI MRI Lumbar Spine Reader (RAI)
1. Table of Acceptance Criteria and Reported Device Performance
Primary Endpoints (Measurements)
| Measurement | MAE Limit (Acceptance Criteria) | Reported MAE (95% CI Upper Bound) | Success |
|---|---|---|---|
| Dural Sac Area (Axial) | 20 mm² | 17.9 mm² | Yes |
| Spinal Canal Area (Axial) | 30 mm² | 24.3 mm² | Yes |
| Lordotic Angle (Sagittal) | 6° | 3.9 ° | Yes |
| Vertebral Body Slippage (Sagittal) | 2 mm | 0.8 mm | Yes |
Secondary Endpoints (Segmentations)
| Anatomical Structure Segmentation | MDC Limit (Acceptance Criteria) | Reported MDC (95% CI Lower Bound) | Success |
|---|---|---|---|
| Artery (Axial) | 0.8 | 0.861 | Yes |
| Disc (Axial) | 0.7 | 0.796 | Yes |
| Disc (Sagittal) | 0.7 | 0.910 | Yes |
| Disc Material Outside IV Space (Axial) | 0.7 | 0.793 | Yes |
| Dural Sac (Axial) | 0.8 | 0.924 | Yes |
| Kidney (Axial) | 0.8 | 0.872 | Yes |
| Ligamentum Flavum (Axial) | 0.7 | 0.736 | Yes |
| Muscle (Axial) | 0.8 | 0.945 | Yes |
| Sacrum (Sagittal) | 0.8 | 0.923 | Yes |
| Spinal Canal (Axial) | 0.8 | 0.941 | Yes |
| Spinal Canal (Sagittal) | 0.8 | 0.865 | Yes |
| Vein (Axial) | 0.8 | 0.815 | Yes |
| Vertebral Arch (Axial) | 0.8 | 0.843 | Yes |
| Vertebral Body (Sagittal) | 0.8 | 0.894 | Yes |
Secondary Endpoints (Measurements)
| Measurement | MAE Limit (Acceptance Criteria) | Reported MAE (95% CI Upper Bound) | Success |
|---|---|---|---|
| Anterior Disc Height (Sagittal) | 2 mm | 1.32 mm | Yes |
| Anterior Vertebral Body Height (Sagittal) | 2 mm | 1.81 mm | Yes |
| Dural Sac Anterior-Posterior Diameter (Axial) | 2 mm | 1.52 mm | Yes |
| Dural Sac Transverse Diameter (Axial) | 2 mm | 1.22 mm | Yes |
| Middle Disc Height (Sagittal) | 2 mm | 1.04 mm | Yes |
| Middle Vertebral Body Height (Sagittal) | 2 mm | 1.33 mm | Yes |
| Posterior Disc Height (Sagittal) | 2 mm | 1.07 mm | Yes |
| Posterior Vertebral Body Height (Sagittal) | 2 mm | 1.72 mm | Yes |
| Spinal Canal Anterior-Posterior Diameter (Axial) | 2 mm | 0.83 mm | Yes |
| Spinal Canal Transverse Diameter (Axial) | 2 mm | 1.85 mm | Yes |
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size: 200 MR image studies for 200 patients.
- Data Provenance: The data was collected from three (3) geographically diverse sites across the U.S. The study was a "standalone software performance study conducted in the U.S." This indicates the data was retrospective as it involved "previously-acquired MR studies" for analysis.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Their Qualifications
- Number of Experts: Five (5) radiologists.
- Qualifications: They were U.S. radiologists (no specific experience years were mentioned, but it's implied they were qualified to interpret spine MRI exams consistent with ACR recommendations, as stated for "qualified users").
4. Adjudication Method for the Test Set
- For Segmentation: For each anatomical structure, the ground truth was established by the per-pixel majority opinion of the five (5) radiologists. Specifically, if at least 3 of the 5 radiologists labeled a pixel as belonging to a particular anatomical structure, the pixel was included; otherwise, it was excluded.
- For Measurement: The ground truth was established by taking the mean of the five (5) radiologists' 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 study was described as a "standalone software performance study."
- Therefore, no effect size for human readers improving with AI vs. without AI assistance was reported.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Was Done
- Yes, a standalone software performance study was conducted. The study compared the RAI software outputs directly to the ground truth established by the five radiologists, without human readers interacting with the AI outputs in a diagnostic setting.
7. The Type of Ground Truth Used
- Expert Consensus: The ground truth was established by the independent assessment and consensus of five (5) U.S. radiologists. For segmentations, it was a majority vote (3 out of 5); for measurements, it was the mean of their individual measurements.
8. The Sample Size for the Training Set
- The document states that "The RAI software machine learning algorithm training and testing data used during the algorithm development... were all independent data sets." However, the specific sample size for the training set is not provided in this document.
9. How the Ground Truth for the Training Set Was Established
- The document mentions that training data was used during algorithm development, but it does not explicitly detail how the ground truth for this training set was established. It only states that the training data and testing data were independent.
§ 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).