Search Results

RadioLens v1.0 software is a medical diagnostic application that displays, processes, stores, and transfers DICOM data, with the exception of mammography. It provides the capability to store images and patient information, perform filtering, digital manipulation, and quantitative measurements. The client software is designed to run on standard personal and business computers.

RadioLens v1.0 includes an optional SpindleX module which is used to analyze potential vertebral body displacements in spine, either absolute or relative, using cervical/lumbar digital X-Rays of spine. A qualified medical practitioner may use the module to semi-automatically identify relevant spine anatomy and calculate vertebral displacement measurements. Measurements are then used to determine severity and location of any spinal ligament injurv leading to subluxation using criteria published in AMA Guides to The Evaluation of Permanent Impairment, 5th & 6th edition.

SpindleX is meant for assistance with analysis of spinal ligament injury in the adult population (>18 years of age) as captured in stress x-rays. SpindleX is not meant for patients that have undergone surgery or other traumatic injury that result in distortion of spinal anatomy.

RadioLens v1.0 is a medical image management and processing system for general human radiology reporting and provides a browser based interface for transfer of DICOM files from the scanning equipment to Synapsica's cloud server, for distribution to authorized users, including Radiologists. It also provides a worklist that keeps track of patient cases that the user needs to work on; a DICOM Viewer that allows users to view and analyze images in patient scan; and a report editor where the Radiologist can prepare the text report that goes back to the patient and referring provider.

RadioLens v1.0 includes an optional SpindleX module. The SpindleX is an artificial intelligence (AI) module that works on digital stress x-rays of the spine and assists with clinical interpretation by a qualified Radiologist by automatically providing multiple relevant measurements in the spine.

SpindleX is meant for assistance with analysis of spinal ligament injury in the adult population (>18 years of age) as captured in stress x-rays. SpindleX is not meant for patients that have undergone surgery or other traumatic injury that result in distortion of spinal anatomy.

The following visualization, quantification, and reporting functionalities are provided by the software:

Visualization

• o 2D image review
• Image navigation tools ●
• Patient worklist ●

Quantitative Analysis
The subject device performs the following anatomical measurements:

• Ruth Jackson's angle,
• Ferguson's angle
• Canal diameter ●
• Displacement of L3 vertebrae ●
• Vertebral offset 0
• Motion Segment Integrity, Translational & Angular .

Reporting
The subject device provides a detailed, objective report for stress of spine following standards set by American Medical Association (AMA). The report is pre-populated with annotated snapshots of digital x-ray scan of the spine generated by SpindleX. Relevant measurements, per the Quantitative Analysis discussed above, are provided along the annotated images. Once the report has been edited and finalized by the Radiologist, it can be downloaded in microsoft word or pdf format.

Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided text:

1. Table of Acceptance Criteria and Reported Device Performance

ICC Categorization:
0.9 Excellent | Cervical:

• Stress Lines (degrees): Rads ICC 0.957 (E), Rads & Model ICC 0.953 (E)
• Canal Diameter (mm): Rads ICC 0.950 (E), Rads & Model ICC 0.955 (E)
• Vertebral Offset Neutral (mm): Rads ICC 0.698 (M), Rads & Model ICC 0.707 (M)
• Angular MSI (degrees): Rads ICC 0.769 (G), Rads & Model ICC 0.781 (G)
• Translational Motion 5th Edition (mm): Rads ICC 0.712 (M), Rads & Model ICC 0.717 (M)
• Translational Motion 6th Edition - Flexion (mm): Rads ICC 0.793 (G), Rads & Model ICC 0.802 (G)
• Translational Motion 6th Edition - Extension (mm): Rads ICC 0.707 (M), Rads & Model ICC 0.716 (M)

Lumbar:

• Stress Lines (degrees): Rads ICC 0.979 (E), Rads & Model ICC 0.968 (E)
• Canal Diameter (mm): Rads ICC 0.826 (G), Rads & Model ICC 0.784 (G)
• Ferguson's Angle (degrees): Rads ICC 0.968 (E), Rads & Model ICC 0.939 (E)
• Integrity of 3rd Lumbar Vertebra (mm): Rads ICC 0.997 (E), Rads & Model ICC 0.996 (E)
• Vertebral Offset Neutral (mm): Rads ICC 0.820 (G), Rads & Model ICC 0.777 (G)
• Angular MSI (degrees): Rads ICC 0.792 (G), Rads & Model ICC 0.771 (G)
• Translational Motion 5th Edition (mm): Rads ICC 0.710 (M), Rads & Model ICC 0.656 (M)
• Translational Motion 6th Edition - Flexion (mm): Rads ICC 0.809 (G), Rads & Model ICC 0.774 (G)
• Translational Motion 6th Edition - Extension (mm): Rads ICC 0.823 (G), Rads & Model ICC 0.769 (G) |

2. Sample size used for the test set and the data provenance

• Sample Size: 600 de-identified stress x-rays (300 cervical and 300 lumbar).
• Data Provenance: Retrospective studies, collected from over 200 institutions in the United States. Each case included flexion, extension, and neutral positions of the spine. The data was independent from the training data.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts

• Number of Experts: Three (3) expert radiologists.
• Qualifications: Not explicitly stated beyond "expert radiologists."

4. Adjudication method for the test set

• The text describes that the 3 expert radiologists analyzed each case with and without the SpindleX module. For the measurement predictions, the "model agreement was equivalent to agreement that radiologists had among themselves," evaluated using ICC scores. This implies that the agreement among the three expert radiologists alone served as a baseline for comparison, and the agreement between radiologists and the model was compared to this baseline agreement. It does not explicitly state a 2+1 or 3+1 adjudication method for establishing a single consensus ground truth prior to comparing with the AI, but rather uses the inter-reader variability as a benchmark.

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

• Yes, a Multi-Reader Multi-Center (MRMC) retrospective study was carried out.
• Effect Size / Improvement with AI: The study evaluated "model agreement was equivalent to agreement that radiologists had among themselves." The ICC tables show the ICC scores "Radiologists ICC" (unaided) and "Rads & Model ICC" (aided). For most measurements, the "Rads & Model ICC" is very close to or slightly higher/lower than the "Radiologists ICC," maintaining the same category (e.g., Excellent, Good, Moderate).
  • For example, in Cervical Stress Lines, Radiologists ICC was 0.957 (E) and Rads & Model ICC was 0.953 (E).
  • For Lumbar Translational Motion 5th Edition, Radiologists ICC was 0.710 (M) and Rads & Model ICC was 0.656 (M).
  • This indicates that the AI module did not significantly "improve" the human readers' agreement beyond their inherent inter-reader agreement, but rather maintained a comparable level of agreement when used in conjunction with the radiologists. The primary claim is that the model's agreement with the radiologists is as good as the radiologists' agreement with themselves, suggesting it seamlessly integrates without degrading performance in terms of inter-observer variability for quantitative measurements.

6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done

• The study design described is a human-in-the-loop performance evaluation ("Three (3) expert radiologists analyzed each case with and without the SpindleX module"). While the model makes predictions, its performance is assessed in terms of its agreement with radiologists, not as a standalone diagnostic tool. The "Rads & Model ICC" evaluates the combined entity. The sensitivities for body part and patient position identification are presented as standalone model performance metrics, suggesting an implicit standalone evaluation for these specific tasks.

7. The type of ground truth used

• For the body part and patient position identification tasks, the ground truth source is not explicitly stated as expert consensus or pathology, but inferred from the radiologists' assessments during the validation study.
• For the quantitative measurements, the ground truth is established by comparing the measurements generated by the model (when used with radiologists) against the agreement observed among the three expert radiologists alone. This effectively uses the expert consensus / inter-reader agreement of the three radiologists as the reference for evaluating the model's performance in achieving "equivalent" agreement.

8. The sample size for the training set

• Training Set Size:
  • 13,452 cervical retrospective studies
  • 6,492 lumbar retrospective studies

9. How the ground truth for the training set was established

• The ground truth for the training set was established by US board-certified Radiologists who pre-marked morphometry points on cervical and lumbar images. These pre-marked points were used to prepare clinical reports for patients and were treated as the ground truth for training the models.

Ask a specific question about this device