(98 days)
The diagnostic ultrasound system and probes are designed to obtain ultrasound images and analyze body fluids.
The clinical applications include: Fetal/Obstetrics, Abdominal, Gynecology, Intra-operative, Small Organ, Neonatal Cephalic, Adult Cephalic, Trans-vaginal, Muscular-Skeletal (Conventional, Superficial), Urology, Cardiac Adult, Cardiac Pediatric, Thoracic, Trans-esophageal (Cardiac) and Peripheral vessel.
It is intended for use by, or by the order of, and under the supervision of, an appropriately trained healthcare professional who is qualified for direct use of medical devices. It can be used in hospitals, private practices, clinics and similar care environment for clinical diagnosis of patients.
Modes of Operation: 2D mode, Color Doppler mode, Power Doppler (PD) mode, M mode, Pulsed Wave (PW) Doppler mode, Continuous Wave (CW) Doppler mode, Tissue Doppler Imaging (TDI) mode, Tissue Doppler Wave (TDW) mode, ElastoScan Mode, Combined modes, Multi-Image mode(Dual, Quad), 3D/4D mode
The V8 / V7 are a general purpose, mobile, software controlled, diagnostic ultrasound system. Their function is to acquire ultrasound data and to display the data as 2D mode, Color Doppler mode, Power Doppler (PD) mode, M mode, Pulsed Wave (PW) Doppler mode, Continuous Wave (CW) Doppler mode, Tissue Doppler Imaging (TDI) mode, Tissue Doppler Wave (TDW) mode, ElastoScan Mode, Combined modes, Multi-Image mode(Dual, Quad), 3D/4D mode. The V8 / V7 also give the operator the ability to measure anatomical structures and offer analysis packages that provide information that is used to make a diagnosis by competent health care professionals. The V8 / V7 have a real time acoustic output display with two basic indices, a mechanical index and a thermal index, which are both automatically displayed.
The document describes the acceptance criteria and study results for several AI-based features integrated into the Samsung V8/V7 Diagnostic Ultrasound System, specifically: NerveTrack, BiometryAssist, ViewAssist, and HeartAssist.
Here's a breakdown of the requested information for each feature:
1. NerveTrack Feature
1.1. Table of Acceptance Criteria and Reported Device Performance
| Validation Type | Definition | Acceptance Criteria | Reported Average | Standard Deviation | 95% Confidence Interval |
|---|---|---|---|---|---|
| Accuracy (%) | Number of correctly detected frames / Total frames with nerve * 100 | ≥ 80% | 90.3 | 4.8 | 88.6 to 92.0 |
| Speed (FPS) | 1000 / Average latency time of each frame (msec) | ≥ 2 FPS | 3.54 | 0.13 | 3.47 to 3.61 |
1.2. Sample Size and Data Provenance for Test Set
- Number of Subjects: 18 (13 Females, 5 Males)
- Number of Images: 2,146
- Data Provenance: All Koreans. Retrospective and Prospective data collection on Samsung ultrasound devices (including V8).
1.3. Number and Qualifications of Experts for Ground Truth (Test Set)
- Number of Experts: 10 anesthesiologists and 5 sonographers.
- Qualifications: All experts had more than 10 years of experience.
1.4. Adjudication Method for Test Set
- One anesthesiologist scanned the ultrasound and drew the initial ground truth.
- Two or more other anesthesiologists and sonographers reviewed and confirmed the ground truth.
- If any mistake was identified during review, it was revised. (This suggests a consensus/adjudication process, effectively 2+1 or more experts for verification after initial labeling).
1.5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done?
- No, an MRMC study comparing human readers with and without AI assistance was not reported for NerveTrack. This study focused on the standalone performance of the AI algorithm.
1.6. If a Standalone (algorithm only) performance was done?
- Yes, the performance data provided (Accuracy and Speed) are for the standalone (algorithm only) performance of the NerveTrack feature.
1.7. Type of Ground Truth Used
- Expert consensus (manual delineation/labeling of nerve locations by multiple experienced anesthesiologists and sonographers).
1.8. Sample Size for Training Set
- Not specified. The document only states that the training data is independent of the test data.
1.9. How the Ground Truth for the Training Set was Established
- Not specified, but it's implied that a similar expert-based process was used, as it states "The training data used for the training of the NerveTrack algorithm are independent of the data used to test the NerveTrack algorithm."
2. BiometryAssist Feature
2.1. Table of Acceptance Criteria and Reported Device Performance
| Validation Type | Definition | Acceptance Criteria | Reported Performance |
|---|---|---|---|
| Segmentation | Dice-score | 0.8 (threshold) | Average Dice-score 0.928 |
| Size Measurement | Circumference Error | ≤ 8% | 8% or less |
| Size Measurement | Distance Error | ≤ 4% | 4% or less |
| Size Measurement | NT Error | ≤ 1mm | 1mm or less |
2.2. Sample Size and Data Provenance for Test Set
- Number of Images: 320 fetal biometry images.
- Data Provenance: Collected at two hospitals. Ethnicity/Country: Americans and Koreans.
- Retrospective/Prospective: Mix of retrospective and prospective data collection. Only V8 images were used for the performance test.
2.3. Number and Qualifications of Experts for Ground Truth (Test Set)
- Number of Experts: Three participating experts (obstetricians) and two sonographers. The entire process was supervised by another obstetrician.
- Qualifications: Obstetricians with more than 20 years of experience, sonographers with more than 10 years of experience, and a supervising obstetrician with more than 25 years of experience; all in fetal cardiology (though the feature is biometry, experts are stated as "in fetal cardiology").
2.4. Adjudication Method for Test Set
- All acquired images were first classified into correct views by three participating experts.
- Corresponding anatomy areas were manually drawn for each image.
- The entire process was supervised (implying a form of consensus/adjudication).
2.5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done?
- No, an MRMC study was not reported. This study assessed the standalone algorithm's performance against expert ground truth.
2.6. If a Standalone (algorithm only) performance was done?
- Yes, the validation was for the deep learning based segmentation algorithm and its measurement capabilities.
2.7. Type of Ground Truth Used
- Expert consensus (manual drawing/delineation of anatomy areas by multiple experienced obstetricians and sonographers).
2.8. Sample Size for Training Set
- Not specified. The data used for training, tuning, and validation are completely separated.
2.9. How the Ground Truth for the Training Set was Established
- Implied to be similar to the test set ground truth establishment: "All acquired images for training, tuning and validation were first classified into the correct views by three participating experts. Afterwards, corresponding anatomy areas were manually drawn for each of the images."
3. ViewAssist Feature
3.1. Table of Acceptance Criteria and Reported Device Performance
| Validation Type | Definition | Acceptance Criteria | Reported Performance |
|---|---|---|---|
| View Recognition | Recognition Accuracy (%) | 89% (threshold) | Average 94.56% |
| Anatomy Annotation (Segmentation) | Dice-score | 0.8 (threshold) | Average Dice-score 0.898 |
3.2. Sample Size and Data Provenance for Test Set
- Number of Images: 680 fetal heart and fetal biometry images.
- Data Provenance: Collected at two hospitals. Ethnicity/Country: Americans and Koreans.
- Retrospective/Prospective: Mix of retrospective and prospective data collection. Only V8 images were used for the performance test.
3.3. Number and Qualifications of Experts for Ground Truth (Test Set)
- Number of Experts: Three participating experts (obstetricians) and two sonographers. The entire process was supervised by another obstetrician.
- Qualifications: Obstetricians with more than 20 years of experience, sonographers with more than 10 years of experience, and a supervising obstetrician with more than 25 years of experience; all in fetal cardiology.
3.4. Adjudication Method for Test Set
- All acquired images were first classified into correct views by three participating experts.
- Corresponding anatomy areas were manually drawn for each image.
- The entire process was supervised (implying a form of consensus/adjudication).
3.5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done?
- No, an MRMC study was not reported. This study assessed the standalone algorithm's performance against expert ground truth.
3.6. If a Standalone (algorithm only) performance was done?
- Yes, the validation was for the deep learning based view recognition and anatomy annotation algorithms.
3.7. Type of Ground Truth Used
- Expert consensus (manual classification of views and delineation of anatomy areas by multiple experienced obstetricians and sonographers).
3.8. Sample Size for Training Set
- Not specified. The data used for training, tuning, and validation are completely separated.
3.9. How the Ground Truth for the Training Set was Established
- Implied to be similar to the test set ground truth establishment: "All acquired images for training, tuning and validation were first classified into the correct views by three participating experts. Afterwards, corresponding anatomy areas were manually drawn for each of the images."
4. HeartAssist Feature
4.1. Table of Acceptance Criteria and Reported Device Performance
| Validation Type | Definition | Acceptance Criteria | Reported Performance |
|---|---|---|---|
| View Recognition (Fetus) | Recognition Accuracy (%) | 89% (threshold) | Average 93.21% |
| View Recognition (Adult) | Recognition Accuracy (%) | 84% (threshold) | Average 98.31% |
| Segmentation (Fetus) | Dice-score | 0.8 (threshold) | Average Dice-score 0.88 |
| Segmentation (Adult) | Dice-score | 0.9 (threshold) | Average Dice-score 0.93 |
| Size Measurement (Fetus) | Area Error | ≤ 8% | 8% or less |
| Size Measurement (Fetus) | Angle Error | ≤ 4% | 4% or less |
| Size Measurement (Fetus) | Circumference Error | ≤ 11% | 11% or less |
| Size Measurement (Fetus) | Diameter Error | ≤ 11% | 11% or less |
| Size Measurement (Adult, B-mode) | PCC value | ≥ 0.8 | Pass |
| Size Measurement (Adult, M-mode) | PCC value | ≥ 0.8 | Pass |
| Size Measurement (Adult, Doppler-mode) | PCC value | ≥ 0.8 | Pass |
4.2. Sample Size and Data Provenance for Test Set
- Number of Images: 888 fetal heart and adult heart images.
- Data Provenance: Collected at five hospitals. Ethnicity/Country: Americans and Koreans.
- Retrospective/Prospective: Mix of retrospective and prospective data collection. Only V8 images were used for the performance test.
4.3. Number and Qualifications of Experts for Ground Truth (Test Set)
- Fetus: Three participating experts (obstetricians) and two sonographers. The entire process was supervised by another obstetrician.
- Qualifications: Obstetrician with more than 20 years of experience, two sonographers with more than 10 years of experience, and a supervising obstetrician with more than 25 years of experience; all in fetal cardiology.
- Adult: Four professionals: two cardiologists and two sonographers.
- Qualifications: Cardiologists with at least 10 years of experience, and sonographers with at least 10 years of experience.
4.4. Adjudication Method for Test Set
- Fetus: All acquired images were first classified into correct views by three participating experts. Corresponding anatomy areas were manually drawn. The entire process was supervised.
- Adult: Experts manually traced the contours of the heart and the signal outline on the images. (Implies consensus/adjudication or a single expert annotation considered ground truth, though multiple experts were involved in the process).
4.5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done?
- No, an MRMC study was not reported. This study assesses the standalone algorithm's performance against expert ground truth.
4.6. If a Standalone (algorithm only) performance was done?
- Yes, the validation was for the deep learning based view recognition, segmentation, and size measurement algorithms.
4.7. Type of Ground Truth Used
- Expert consensus (manual classification of views, delineation of anatomy areas, and measurement of values by multiple experienced obstetricians/cardiologists and sonographers). For adult measurements, the "cardiologist's measurements" were defined as ground truth.
4.8. Sample Size for Training Set
- Not specified. The data used for training, tuning, and validation are completely separated.
4.9. How the Ground Truth for the Training Set was Established
- Implied to be similar to the test set ground truth establishment:
- Fetus: "All acquired images for training, tuning and validation were first classified into the correct views by three participating experts. Afterwards, corresponding anatomy areas were manually drawn for each of the image."
- Adult: Similar process of experts manually tracing contours and signal outlines.
§ 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.