Search Results

The ECHELON Synergy System is an imaging device and is intended to provide the physician with physiological and clinical information, obtained non-invasively and without the use of ionizing radiation. The MR system produces transverse, coronal, sagittal, oblique, and curved cross sectional images that display the internal structure of the head, body, or extremities. The images produced by the MR system reflect the spatial distribution of protons (hydrogen nuclei) exhibiting magnetic resonance. The NMR properties that determine the image appearance are proton density, spinlattice relaxation time (TI), spin-spin relaxation time (T2) and flow. When interpreted by a trained physician, these images provide information that can be useful in diagnosis determination.

Anatomical Region: Head, Body, Spine, Extremities
Nucleus excited: Proton

Diagnostic uses:

• · TI, T2, proton density weighted imaging
• · Diffusion weighted imaging
• · MR Angiography
• · Image processing
• · Spectroscopy
• · Whole Body

The ECHELON Synergy is a Magnetic Resonance Imaging System that utilizes a 1.5 Tesla superconducting magnet in a gantry design.

The provided document is a 510(k) summary for the FUJIFILM Healthcare Corporation's ECHELON Synergy MRI System. This document asserts substantial equivalence to a predicate device and primarily focuses on technical characteristics and adherence to standards rather than detailed performance studies with acceptance criteria for a diagnostic aid.

Here's an analysis of the acceptance criteria and study information derived from the document:

1. A table of acceptance criteria and the reported device performance:

The document doesn't explicitly state quantitative acceptance criteria in terms of diagnostic performance metrics (e.g., sensitivity, specificity, AUC) because it's a 510(k) submission for an MRI system with an added coil, not a diagnostic algorithm. The acceptance criteria for the added Breast Coil 17 are implicitly tied to the performance and safety standards of the predicate device (ECHELON Synergy V10.0 K233687).

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

• Sample size for test set: Not explicitly stated as a number of cases or patients. The document mentions "Clinical image examples."
• Data provenance: Not explicitly stated (e.g., country of origin, retrospective or prospective). It only states that "Clinical images were collected and analyzed."

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

• Number of experts: One radiologist.
• Qualifications of experts: A "radiologist" validated the clinical images. No further details on experience level are provided.

4. Adjudication method for the test set:

• Adjudication method: None mentioned beyond a single radiologist's validation of image quality for clinical use.

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:

• MRMC study: No, an MRMC comparative effectiveness study was not explicitly mentioned or implied. This submission is for an MRI system with an added coil, not an AI-powered diagnostic algorithm.

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

• Standalone performance: Not applicable. This device is an MRI system, not a standalone AI algorithm. The performance evaluation focused on the technical aspects and image quality of the MRI machine and its new coil.

7. The type of ground truth used:

• Type of ground truth: Expert opinion (a single radiologist's validation of "acceptable image quality for clinical use"). This is tied to the demonstrative aspect of clinical image examples, rather than a definitive diagnostic truth for a disease state.

8. The sample size for the training set:

• Sample size for training set: Not applicable. This document is about a hardware modification (an added coil) to an existing MRI system. It does not involve machine learning models that require training sets in the conventional sense.

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

• How ground truth for training set was established: Not applicable, as there is no mention of a training set or machine learning model.

Ask a specific question about this device

The ECHELON Synergy System is an imaging device and is intended to provide the physician with physiological and clinical information, obtained non-invasively and without the use of ionizing radiation. The MR system produces transverse, coronal, sagittal, oblique, and curved cross-sectional images that display the internal structure of the head, body, or extremities. The images produced by the MR system reflect the spatial distribution of protons (hydrogen nuclei) exhibiting magnetic resonance. The NMR properties that determine the image appearance are proton density, spin-lattice relaxation time (T1), spin-spin relaxation time (T2) and flow. When interpreted by a trained physician, these images provide information that can be useful in diagnosis determination.

The ECHELON Synergy is a Magnetic Resonance Imaging System that utilizes a 1.5 Tesla superconducting magnet in a gantry design. The design was based on the ECHELON OVAL V6.0A MRI system. The ECHELON Synergy has been designed to enhance clinical utility as compared to the ECHELON OVAL V6.0A by taking advantage of open architecture.

The provided document, K223426, is a 510(k) premarket notification for the FUJIFILM Healthcare Corporation's ECHELON Synergy MRI system. This submission primarily focuses on demonstrating substantial equivalence to a predicate device (ECHELON OVAL V6.0A MRI system, K172110) rather than presenting a detailed performance study with explicit acceptance criteria for an AI/ML powered device as typically required for novel AI products.

However, the document mentions several new features powered by Machine Learning (ML), specifically Deep Learning Reconstruction (DLR), AutoClip, AutoPose Spine, AutoPose Shoulder, and AutoPose Knee. For DLR, some form of evaluation was performed. For AutoClip and AutoPose functions, performance comparisons were made against manual operations.

Based on the provided text, a comprehensive table of acceptance criteria and reported device performance, as one would expect for a dedicated AI/ML device approval, is not explicitly stated with numerical thresholds. The evaluations are largely qualitative or comparative to existing methods.

Below is an attempt to extract the closest information to your request, specifically focusing on the DLR, AutoClip, and AutoPose functions, as they are the only "AI/ML powered" components mentioned with specific evaluations.

1. Table of Acceptance Criteria and Reported Device Performance

As explicit numerical acceptance criteria are not provided for the AI/ML components, the table below consolidates the stated evaluative goals and findings from the "Summary of Clinical Testing" section.

• SNR: Equivalent or better in 81 out of 81 cases.
• Sharpness: Equivalent or better in 80 out of 81 cases.
• Lesion Conspicuity: Equivalent or better in 45 out of 45 cases (with pathology).
• Overall Image Quality: Equivalent or better in all cases.
  Motion Artifact Handling: Rated as better or equivalent image quality in all 3 image pairs with motion artifacts, indicating DLR did not significantly change their appearance.
  Shorter Scan Time Efficacy: DLR images with shorter scan times were rated "acceptable for routine examinations" in all 18 cases.
  Resolution Improvement: High-resolution DLR images were rated "better or equivalent" image quality in all cases compared to low-resolution conventional images. |
  | AutoClip | Performance Equivalence: Performance should be "substantially equivalent" to manual clipping. | Confirmed that the performance of AutoClip was "substantially equivalent to that of manual clipping." |
  | AutoPose (Spine, Shoulder, Knee) | Efficiency Improvement/Equivalence: Should reduce time and number of steps in slice positioning compared to manual, or at least show the "same time and number of steps." | Spine, Shoulder, and Knee:
• Many cases were able to reduce the time and number of steps in slice positioning compared to manual.
• Remaining cases showed the same time and number of steps as manual slice positioning. |

2. Sample Sizes Used for the Test Set and Data Provenance

• Deep Learning Reconstruction (DLR):

  • Number of cases: 110 cases for DLR image quality evaluation (including 81 cases for SNR/sharpness/overall IQ, 45 cases with pathology for lesion conspicuity, 3 cases for motion artifacts, and 18 cases for shorter scan time evaluation). The exact breakdown per sub-analysis is specified.
  • Data Provenance: ECHELON OVAL, ECHELON Smart, and ECHELON Synergy MRI systems (all FUJIFILM Healthcare Corporation 1.5T MRI systems). Data acquired at "FUJIFILM Healthcare Corporation and clinical site."
  • Subject Type: Healthy volunteer and patient.
  • Anatomical Coverage: Head, Spine, Cardiac, Breast, Abdomen, Pelvis, Shoulder, Wrist, Knee, Ankle.

• AutoClip:

  • Number of cases: 40 cases.
  • Data Provenance: ECHELON Synergy MRI system (FUJIFILM Healthcare Corporation 1.5T MRI system). Data acquired at "FUJIFILM Healthcare Corporation."
  • Subject Type: Japanese healthy volunteers.
  • Anatomical Coverage: Brain (using 3D TOF, 3D Soft TOF scan sequences).

• AutoPose (Spine, Shoulder, Knee):

  • Number of cases: Spine (146 cases), Shoulder (48 cases), Knee (38 cases).
  • Data Provenance: ECHELON Synergy MRI system (FUJIFILM Healthcare Corporation 1.5T MRI system). Data acquired at "FUJIFILM Healthcare Corporation."
  • Subject Type: Japanese healthy volunteers.
  • Anatomical Coverage: Spine, Shoulder, Knee.

Note: The document does not explicitly state if the data was retrospective or prospective. Given the nature of performance testing within a company and potentially a clinical site, it could be a mix or internal prospective collection, but it's not specified. The country of origin for the "clinical site" data is also not explicitly stated beyond "Japanese healthy volunteers" for AutoClip/AutoPose, implying at least part of the data is from Japan for those features.

3. Number of Experts Used to Establish Ground Truth for the Test Set and Qualifications

• Deep Learning Reconstruction (DLR):

  • Number of Experts: Three (3) US certified radiologists.
  • Qualifications: "US certified radiologists." No specific years of experience or subspecialty are provided, beyond their certification.

• AutoClip & AutoPose (Spine, Shoulder, Knee):

  • Number of Experts: Not specified as "experts" establishing ground truth, but rather "certified radiological technologists" performed the performance comparison/evaluation. The ground truth for performance was implicitly "manual operation" by these technologists. Their qualifications are listed as "certified radiological technologists."

4. Adjudication Method for the Test Set

• Deep Learning Reconstruction (DLR): The document states "Readers compared pairs of DLR images and conventional images (without DLR) for each case to evaluate image quality of DLR images." It does not specify an explicit adjudication method (e.g., 2+1, 3+1). It merely presents the results as derived from the collective evaluation of the three radiologists. It's unclear if consensus was required, or if individual ratings were aggregated.

• AutoClip & AutoPose: The evaluation was done by "certified radiological technologists" comparing against manual operation. No formal adjudication process is described.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, and Effect Size

• Deep Learning Reconstruction (DLR): A study involving multiple readers (3 US certified radiologists) and multiple cases (110 cases in total for DLR evaluation) was performed, which aligns with the spirit of an MRMC study. However, it's not explicitly labeled as such, and the methodology primarily focuses on qualitative comparison of image quality between DLR and conventional images rather than a comparative effectiveness study of human reader diagnostic performance with vs. without AI assistance for a specific diagnostic task.

  • Effect Size of Human Reader Improvement: This type of effect size (e.g., AUC uplift) is not reported. The study focused on assessing image quality attributes and acceptability for routine examinations from the DLR images themselves, as perceived by radiologists, not on how DLR assistance changes a radiologist's diagnostic accuracy or efficiency on a specific clinical task. The evaluation was primarily about the AI's impact on image characteristics, not human diagnostic performance.

• AutoClip & AutoPose: These evaluations were focused on the efficiency and equivalence of the automated process compared to manual operation, as assessed by technologists. They were not MRMC studies designed to measure impact on human readers' diagnostic effectiveness.

6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) was done

• The document implies a standalone assessment of the DLR output quality (SNR, sharpness, etc.) against conventional images, as rated by radiologists. The AutoClip and AutoPose functions are also inherently standalone algorithms that automate tasks, with their performance evaluated against manual methods. However, no formal "standalone performance study" with typical metrics like sensitivity, specificity, or AUC for a diagnostic task is presented for these AI/ML components in isolation. The evaluation focuses on product-level performance and usability.

7. The Type of Ground Truth Used

• Deep Learning Reconstruction (DLR): The ground truth for evaluation was expert consensus/opinion (or individual expert assessment) of the image quality attributes (SNR, sharpness, lesion conspicuity, overall image quality) when comparing DLR images to conventional images. The underlying "ground truth" for the cases themselves (e.g., presence of pathology) would presumably come from standard clinical diagnostic reports or other confirmed findings, but the DLR study's focus was on image quality as assessed by experts.

• AutoClip & AutoPose: The ground truth for these functions was the manual operation by certified radiological technologists. The evaluation aimed to determine if the automated function delivered equivalent or better performance (in terms of results and/or efficiency) compared to the human-performed task.

8. The Sample Size for the Training Set

• The document does not provide any details on the sample size used for the training set for DLR, AutoClip, or AutoPose. This information is typically proprietary and not included in 510(k) summaries unless specifically requested by the FDA or deemed critical for demonstrating substantial equivalence for a novel AI/ML device.

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

• Similar to the training set sample size, the document does not provide any details on how the ground truth for the training set was established for DLR, AutoClip, or AutoPose.

Ask a specific question about this device