LogoFDA 510(k) Search
K Number
K231965
Device Name
Voluson Expert 22, Voluson Expert 20, Voluson Expert 18
Manufacturer
GE Medical Systems Ultrasound and Primary Care Diagnostics,
Date Cleared
2023-10-30

(119 days)

Product Code
IYN,ITX,IYO
Regulation Number
892.1550
Type
Traditional
Panel
RA
Reference & Predicate Devices
K230346,K220358
AI/MLSaMDIVD (In Vitro Diagnostic)TherapeuticDiagnosticis PCCP AuthorizedThirdpartyExpeditedreview
Intended Use

The device is a general purpose ultrasound system intended for use by qualified and trained healthcare professionals. Specific clinical applications remain the same as previously cleared: Fetal/OB; Abdominal (including GYN, pelvic and infertility monitoring/follicle development); Pediatric; Small Organ (breast, testes, thyroid etc.); Neonatal and Adult Cephalic; Cardiac (adult and pediatric); Musculo-skeletal Conventional and Superficial; Vascular; Transvaginal (including GYN); Transrectal
Modes of operation include: B, M, PW Doppler, CW Doppler, Color M Doppler, Power Doppler, Power Doppler, Harmonic Imaging, Coded Pulse, 3D/4D Imaging mode, Elastography, Shear Wave Elastography and Combined modes: B/M, B/Color, B/PWD, B/Color/PWD, B/Power/ PWD, B/Elastography. The Voluson™ Expert 20, Voluson™ Expert 22 is intended to be used in a hospital or medical clinic.

Device Description

The systems are full-featured Track 3 ultrasound systems, primarily for general radiology use and specialized for OB/GYN with particular features for real-time 3D/4D acquisition. They consist of a mobile console with keyboard control panel; color LCD/TFT touch panel, color video display and optional image storage and printing devices. They provide high performance ultrasound imaging and analysis and have comprehensive networking and DICOM capability. They utilize a variety of linear, curved linear, matrix phased array transducers including mechanical and electronic scanning transducers, which provide highly accurate real-time three-dimensional imaging supporting all standard acquisition modes.

AI/ML Overview

Here's an analysis of the acceptance criteria and supporting studies for the device, based on the provided text:

Acceptance Criteria and Device Performance

The document describes several AI features: SonoAVC2.0, SonoPelvicFloor2.0, Fibroid Mapping, and SonoLyst/SonoLyst Live.

SonoAVC2.0 is explicitly stated to have migrated from a previously cleared device (K230346, Voluson SWIFT/SWIFT+) as "Auto Caliper," implying its performance has already met acceptance criteria in that context. No new specific acceptance criteria or performance metrics are provided for SonoAVC2.0 in this document.

SonoPelvicFloor2.0

Acceptance CriteriaReported Device Performance
Success rate should be 70% or higher on datasets marked as "Good" in Image/Cine Quality (IQ) assessment.96% accuracy on Good IQ datasets.
Success rate should be 60% or higher on datasets marked as "Challenging" in Image/Cine Quality (IQ) assessment.93% accuracy on Challenging IQ datasets.

Fibroid Mapping

Acceptance CriteriaReported Device Performance (Primary Verification)Reported Device Performance (Secondary Evaluation)
Success rate of each AI component (uterus, endometrium, and fibroid segmentation) should be 70% or higher. (Qualitative assessment)Uterus: 95%Uterus: 100%
Endometrium: 89%Endometrium: 91%
Fibroids: 88%Fibroids: 88%

Fibroid Mapping (Quantitative Evaluation - DICE Scores when successful)

Acceptance CriteriaReported Device Performance
Implicit: High accuracy of algorithms and correlation between DICE scores and qualitative assessment by independent clinical experts. (No specific numerical threshold given as acceptance criterion, but high DICE scores are presented as evidence of high accuracy when successful.)Uterus: 0.89 ± 0.03
Endometrium: 0.70 ± 0.18
Fibroids: 0.70 ± 0.13

SonoLyst/SonoLyst Live (2nd Trimester Features)

Acceptance CriteriaReported Device Performance
Average success rate of SonoLyst 2nd Trimester IR and X and overall traffic light accuracy is 80% or higher.Meets criteria (specific percentages not provided, but stated as "met").

Study Details

SonoPelvicFloor2.0

  1. Sample Size and Data Provenance (Test Set):

    • Total Volumes: 94
    • Distribution by Systems: Voluson Expert 22 (82), Voluson E8 (8), Voluson E6 (4)
    • Distribution by Probes: RAB6-D (41), RIC5-9-D (21), RM7C (32)
    • Distribution by Countries: Austria (36), U.S.A (47), South Africa (4), Czech Republic (7)
    • Data Provenance: Retrospective. Data is "provided by external clinical partners who de-identified the data before sharing it with us." Original data is collected in the form of 4D volume Cines, preserving flexibility for retrospective re-processing. The data covers multiple geographical sites.

  2. Number and Qualifications of Experts for Ground Truth (Test Set):

    • The verification was "performed by clinical experts" and involved an "internal clinical expert" for assessment.
    • Specific numbers or qualifications (e.g., years of experience, specialty) of these experts are not explicitly stated in the provided text.

  3. Adjudication Method (Test Set):

    • Not explicitly described as a formal adjudication method (e.g., 2+1). The "internal clinical expert" documents the assessment.

  4. Multi Reader Multi Case (MRMC) Comparative Effectiveness Study:

    • No MRMC comparative effectiveness study is mentioned for this feature. Performance is reported as standalone accuracy.

  5. Standalone Performance:

    • Yes, the reported accuracy (96% on good IQ, 93% on challenging IQ) is for the algorithm's standalone performance.

  6. Type of Ground Truth:

    • "Internal clinical expert" assessment based on a specific workflow (observing automated placement of MHD plane and assessing success/failure). This would constitute expert consensus/judgment on the algorithm's output.

  7. Sample Size for Training Set:

    • Volumes: 983 distinct volumes
    • Patients: 616 individual patients
    • Images (generated): 8847 images (3 slices from each volume with three rotation angles per image)
    • Distribution by Systems: V730 (116), Voluson E10 (482), Voluson E6 (21), Voluson E8 (90), Voluson P8 (274)
    • Distribution by Probes: RAB 4-8L (116), RAB2-6-RS (38), RAB6-D (111), RIC5-9A-RS (236), RIC5-9-D (8), RM6C (474)
    • Distribution by Countries: Australia (116), Austria (8), Belgium (465), Czech Republic (100), Japan (236), Italy (37), South Africa (21)

  8. How Ground Truth for Training Set was Established:

    • Not explicitly detailed for the training set within the provided text. However, it states that "Data used for both training and validation has been collected across multiple geographical sites" and "Data is provided by external clinical partners who de-identified the data." This implies that clinical experts or medical professionals initially generated the data, which was then de-identified. The process of labeling or establishing "ground truth" (e.g., manual annotations of pelvic floor structures) on these 983 volumes for training is not described.

Fibroid Mapping

  1. Sample Size and Data Provenance (Test Set/Verification):

    • Primary Verification:
      • Total Volumes: 78
      • Systems: Voluson Expert Series
      • Probes: RIC5-9-D, RIC6-12-D
      • Countries: Egypt, Greece, Korea, Philippines, Romania, USA.
    • Secondary Evaluation (Qualitative and Quantitative):
      • Total Volumes: 74
      • Systems: Voluson Expert Series
      • Probes: RIC5-9-D, RIC10-D
      • Countries: Austria, Egypt, Germany, Greece, Korea, Philippines, Romania, USA.
    • Data Provenance: Retrospective. Data "provided by external clinical partners who de-identified the data before sharing it with us." Original data is collected in 3D/4D volumes.

  2. Number and Qualifications of Experts for Ground Truth (Test Set):

    • Verification "performed by clinical experts."
    • "Independent clinical experts" for qualitative assessment and correlation with DICE scores.
    • "Internal and external clinical experts" were consulted for defining requirements for success rates.
    • Specific numbers or qualifications (e.g., years of experience, specialty) of these experts are not explicitly stated in the provided text.

  3. Adjudication Method (Test Set):

    • Not explicitly described as a formal adjudication method. Clinical experts perform the qualitative assessment and judge if segmentation is "satisfying." If not, they may return to an earlier step or activate manual segmentation.

  4. Multi Reader Multi Case (MRMC) Comparative Effectiveness Study:

    • No MRMC comparative effectiveness study is mentioned for this feature. Performance is reported as a standalone success rate and DICE scores.

  5. Standalone Performance:

    • Yes, the reported success rates and DICE scores are for the algorithm's standalone performance in segmentation.

  6. Type of Ground Truth:

    • Qualitative: "Clinical expert" assessment of whether the segmentation is "satisfying."
    • Quantitative: DICE scores calculated against what is presumably expert-annotated ground truth segmentations, which are then correlated with the qualitative expert assessment. The explicit method for defining this precise quantitative ground truth (e.g., expert manual segmentation per pixel) is not detailed for the test set.

  7. Sample Size for Training Set:

    • Unique Volumes: 779
    • Augmented Volumes: 7790

  8. How Ground Truth for Training Set was Established:

    • "We collected training data of the following types from hospital centers as well as private practices. We provided a dedicated data collection protocol explaining the required data needs to each provider." This implies a structured collection, but the actual process of defining the ground truth for segmentation (e.g., manual outlining of fibroids, uterus, endometrium by experts on each training image) is not explicitly described.

SonoLyst/SonoLyst Live (2nd Trimester Features) Improvement

  1. Sample Size and Data Provenance (Test Set):

    • Quantitative Evaluation for SonoLyst 2nd Trimester IR: 42102 images
    • Quantitative Evaluation for SonoLyst 2nd Trimester X: 10424 images
    • Quantitative Evaluation for SonoLystLive 2nd Trimester: 5666 images
    • Data Provenance: Retrospective. Data collected from routine clinical practice in hospitals and private practice during mid-trimester examinations. Countries include UK, Austria, India, and USA.

  2. Number and Qualifications of Experts for Ground Truth (Test Set):

    • Sonographer Panel: 5 sonographers were used to review sorted and graded images for accuracy.
    • Initial Curation: A "single sonographer" initially curated (sorted and graded) the images.
    • Specific qualifications (e.g., years of experience) for these sonographers are not explicitly stated.

  3. Adjudication Method (Test Set):

    • Majority View: For images where the initial sonographer's sorting differed from the algorithm's initial classification, a 5-sonographer review panel determined the sorting accuracy, resulting in reclassification based on the "majority view of the panel."
    • For graded images, the 5-sonographer panel reviewed where they differed from the initial ground truth to determine grading accuracy. This is a form of 5-reader consensus/adjudication.

  4. Multi Reader Multi Case (MRMC) Comparative Effectiveness Study:

    • The verification of SonoLystLive 2nd Trimester features is based on "the average agreement between a sonographer panel and the output of the algorithm regarding Traffic light quality." While this involves multiple readers assessing the algorithm's output, it's described more as a validation of the algorithm's performance against human judgment rather than a comparative effectiveness study showing human reader improvement with AI assistance. No effect size of human improvement with AI vs. without is provided.

  5. Standalone Performance:

    • Yes, the reported average success rate of 80% or higher for SonoLyst 2nd Trimester IR and X and overall traffic light accuracy is a standalone measure of the algorithm's performance in sorting and grading.

  6. Type of Ground Truth:

    • Expert Consensus: A "sonographer review panel" (5 sonographers) provided the ground truth for both sorting and grading accuracy, based on a majority view where discrepancies arose with an initial sonographer's assessment.

  7. Sample Size for Training Set:

    • Total Images: 2.2 million
    • Total Cine Loops: 2570
    • Data collected across multiple geographical sites using different systems (GE Voluson V730, E6, E8, E10, Siemens Acuson S2000 and Hitachi Aloka).

  8. How Ground Truth for Training Set was Established:

    • Not explicitly described for the training set in this section. General statement that "Data used for both training and validation has been collected across multiple geographical sites." The detailed "truthing process" mentioned focuses on the verification data, not the much larger training set. It's common for training data to be annotated by experts, but the specifics are not provided here.

§ 892.1550 Ultrasonic pulsed doppler imaging system.

(a)
Identification. An ultrasonic pulsed doppler imaging system is a device that combines the features of continuous wave doppler-effect technology with pulsed-echo effect technology and is intended to determine stationary body tissue characteristics, such as depth or location of tissue interfaces or dynamic tissue characteristics such as velocity of blood or tissue motion. This generic type of device may include signal analysis and display equipment, patient and equipment supports, component parts, and accessories.(b)
Classification. Class II.