Intended Use / Indications for Use

EPIQ:

The intended use of Philips EPIQ series diagnostic ultrasound systems is diagnostic ultrasound imaging and fluid flow analysis of the human body, with the following indications for use:

Abdominal, Cardiac Adult, Cardiac other (Fetal), Cardiac Pediatric, Cerebral Vascular, Cephalic (Adult), Cephalic (Neonatal), Fetal/Obstetric, Gynecological, Intra-cardiac Echo, Intra-luminal, Intraoperative (Vascular), Intraoperative (Cardiac), Musculoskeletal (Conventional), Musculoskeletal (Superficial), Ophthalmic, Other: Urology, Pediatric, Peripheral Vessel, Small Organ (Breast, Thyroid, Testicle), Transesophageal (Cardiac), Transrectal, Transvaginal, Lung.

Modes of operation include: B Mode, M Mode, PW Doppler, CW Doppler, Color Doppler, Color M Mode, Power Doppler, and Harmonic Imaging.

The clinical environments where Philips EPIQ diagnostic ultrasound systems can be used include clinics, hospitals, and clinical point-of-care for diagnosis of patients.

When integrated with Philips EchoNavigator, the systems can assist the interventionalist and surgeon with image guidance during treatment of cardiovascular disease in which the procedure uses both live X-ray and live echo guidance.

The systems are intended to be installed, used, and operated only in accordance with the safety procedures and operating instructions given in the product user information. Systems are to be operated only by appropriately trained healthcare professionals for the purposes for which they were designed. However, nothing stated in the user information reduces your responsibility for sound clinical judgement and best clinical procedure.

Affiniti:

The intended use of the Affiniti Series Diagnostic Ultrasound Systems is diagnostic ultrasound imaging and fluid flow analysis of the human body with the following Indications for Use:

Abdominal, Cardiac Adult, Cardiac other (Fetal), Cardiac Pediatric, Cerebral Vascular, Cephalic (Adult), Cephalic (Neonatal), Fetal/Obstetric, Gynecological, Intraoperative (Vascular), Intraoperative (Cardiac), Musculoskeletal (Conventional), Musculoskeletal (Superficial), Other: Urology, Pediatric, Peripheral Vessel, Small Organ (Breast, Thyroid, Testicle), Transesophageal (Cardiac), Transrectal, Transvaginal, Lung.

Modes of operation include: B Mode, M Mode, PW Doppler, CW Doppler, Color Doppler, Color M Mode, Power Doppler, and Harmonic Imaging.

The clinical environments where the Affiniti Diagnostic Ultrasound Systems can be used include Clinics, Hospitals, and clinical point-of-care for diagnosis of patients.

The systems are intended to be installed, used, and operated only in accordance with the safety procedures and operating instructions given in the product user information. Systems are to be operated only by appropriately trained healthcare professionals for the purposes for which they were designed. However, nothing stated in the user information reduces your responsibility for sound clinical judgment and best clinical procedure.

The purpose of this Special 510(k) Pre-Market Notification is to expand the Smart View Select (SVS) software application onto the Affiniti Series Diagnostic Ultrasound Systems and to modify the Smart View Select software application onto both the EPIQ Series Diagnostic Ultrasound Systems.

Smart View Select is an automated software feature that assists the user in selection of images for analysis with the existing Philips AutoStrain LV or 2D Auto LV application in Adult Echo Transthoracic (TTE) examination. This feature automatically classifies each acquired image by view and selects an appropriate set of images for left ventricle (LV) analysis. The classification is based on a Deep Learning AI interface engine; the selection is a non-AI algorithm that considers the view classification and image depth to select the optimal set of images.

No hardware changes to the EPIQ or Affiniti systems are required when using SVS, and existing, cleared Philips TTE transducers are used with these software applications.

The SVS v2 feature is supported by all EPIQ models running software version 13.0 or higher including EPIQ CVx/CVxi, EPIQ Elite Advanced, EPIQ Elite, EPIQ 7, EPIQ 5. The SVS v2 feature is supported by the Affiniti models running software version VM13.0 or higher, including Affiniti CVx, Affiniti 70, Affiniti 50, and Affiniti 30. The SVS and SWM software features are associated with the cardiac adult indication.

The provided document details the 510(k) clearance for Philips Ultrasound's EPIQ and Affiniti Series Diagnostic Ultrasound Systems with the Smart View Select (SVS) v2 software. The SVS feature is an automated software that assists users in selecting images for Left Ventricle (LV) analysis using existing Philips AutoStrain LV or 2D Auto LV applications in Adult Echo Transthoracic (TTE) examinations. It classifies acquired images by view using a Deep Learning AI interface and then uses a non-AI algorithm to select an optimal set of images.

Here's an analysis of the acceptance criteria and the study that proves the device meets them:

1. Table of Acceptance Criteria and Reported Device Performance

The study aimed to demonstrate agreement between LV analysis outputs (Ejection Fraction - EF, and Global Longitudinal Strain - GLS) derived from manually selected clips (ground truth) and automatically selected clips by the SVS software.

2. Sample Size and Data Provenance

• Test Set Sample Size: The exact number of patients or cases in the test set is not explicitly stated as a single number. However, the demographic characteristics section indicates n=71 for various demographic variables like Sex, Age, Height, Weight, BMI, Race, LV systolic function, RWMA, Known CAD, Previous reported MI Location, and LV Hypertrophy, suggesting the test dataset comprised data from at least 71 unique patients.
• Data Provenance: The data was collected from a US-based medical center. The study was a retrospective data analysis.

3. Number of Experts and Qualifications for Ground Truth

• Number of Experts: Three (3) reviewers (clinical experts) participated in establishing the ground truth.
• Qualifications of Experts: They are referred to as "clinical experts." No further specific qualifications (e.g., years of experience, specific board certifications like radiologist/cardiologist) are provided in the document.

4. Adjudication Method for the Test Set

The adjudication method used to establish ground truth was consensus/averaging. For each output (GLS and EF), the average across the three reviewers was used as the ground truth.

5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

No Multi-Reader Multi-Case (MRMC) comparative effectiveness study was described. The study focused on comparing the AI-selected clips' performance (GLS and EF) against the ground truth established by human experts using manually selected clips. There is no information provided regarding how much human readers improve with AI vs. without AI assistance. The study evaluates the AI's ability to select optimal clips, a step that precedes human analysis with AutoStrain LV.

6. Standalone Performance Study (Algorithm Only)

Yes, a standalone performance study was done. The performance of the SVS software was evaluated solely based on its ability to automatically select appropriate clips, and then the subsequent, un-edited evaluation of GLS and EF through AutoStrain LV using these SVS-selected clips was compared against the human-selected clips. This represents the algorithm's performance in its specific task (clip selection) without direct human intervention in the selection process.

7. Type of Ground Truth Used

The ground truth used was expert consensus. Specifically, it was the average of GLS and EF measurements obtained from clips manually selected and subsequently semi-automatically processed by three clinical experts using the AutoStrain LV software.

8. Sample Size for the Training Set

The document does not provide the sample size used for the training set of the deep learning AI model for image classification. It only states that "the neural network of the subject device was revised as described in attachment 002," implying updates to an existing model rather than a completely new one, but no training data details are given.

9. How Ground Truth for Training Set was Established

The document does not explicitly state how the ground truth for the training set was established. It mentions that the "classification is based on a Deep Learning AI interface engine," but details regarding the training data annotation and ground truth establishment are not provided within the excerpt.

Ask a specific question about this device

L Vivo platform is intended for non-invasive processing of ultrasound images to detect, measure, and calculate relevant medical parameters of structures and function of patients with suspected disease.

The LVivo System analyzes echocardiographic patient examination DICOM movies for Global ejection fraction (EF) evaluation. EF is evaluated using two orthogonal planes, four-chamber (4CH) and two-chamber (2CH) views, to provide fully automated analyses of LV function from the echo examination movies. It also has the ability to measure strain and to evaluate the Right Ventricle and well as to measure the bladder.

This document describes the acceptance criteria and study results for DiA Imaging Analysis Ltd's LVivo Software Application, specifically focusing on its LVivoRV (Right Ventricular) and LVivo Bladder modules.

1. Table of Acceptance Criteria and Reported Device Performance

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

• LVivoRV: The exact sample size for the test set is not explicitly stated. The study mentions that examinations were collected over a period of 22 months retrospectively. It includes RV clips from 4CH and 4CH modified views that had 2-3 stable recorded beats.

  • Data Provenance: Retrospective, single-center study. The data was retrieved from PACS systems available on-site at the study center.

• LVivo Bladder: 226 bladder images (113 pairs of transverse and longitudinal views) were included.

  • Data Provenance: Retrospective, single-center study. Examinations were retrieved from PACS systems available in Terem's clinic. The data was collected as part of routine abdominal examinations.

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

• LVivoRV:

  • FAC, EDA, ESA: Ground truth was established by two sonographers who manually traced the ED and ES frames. Their qualifications are not explicitly detailed beyond being "sonographers."
  • TAPSE: Ground truth was established by a sonographer (using M-Mode) and an echo cardiologist (using VVI).
  • S': Ground truth was established by a sonographer (using M-Mode) and an echo cardiologist (using VVI).
  • RV Strain: Ground truth was established by an echo cardiologist (using VVI).
  • Overall Qualification Level: Sonographers and echo cardiologists are qualified medical professionals routinely performing these measurements.

• LVivo Bladder:

  • Ground truth was established by one expert sonographer who performed manual measurements of bladder volume (D1, D2, D3) from the trans and long views. The manual measurements were considered the reference/ground truth.

4. Adjudication Method for the Test Set

• LVivoRV:

  • For EDA, ESA, and FAC, the LVivoRV's automated values were compared to the average of the values obtained manually by the two sonographers. This implies a form of consensus or averaging method.
  • For TAPSE and S', measurements from a sonographer (M-Mode) and an echo cardiologist (VVI) were used for comparison, but it's not explicitly stated if there was an adjudication for a single ground truth value when both measurements were available. However, the report compares LVivoRV to "manual measurement using M-Mode" and also notes correlation between VVI and M-Mode, suggesting separate comparisons rather than a combined adjudication.

• LVivo Bladder:

  • The manual measurements performed by the single expert sonographer served as the ground truth. There was no explicit adjudication among multiple experts since only one expert provided the manual measurements for the validation.

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

• No MRMC comparative effectiveness study, evaluating human readers with and without AI assistance, was reported in this document. The studies were focused on the standalone performance of the LVivoRV and LVivo Bladder modules against manual measurements or established methods.

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

• Yes, standalone performance studies were done for both LVivoRV and LVivo Bladder.
  • LVivoRV: The algorithm's output was generated "by an automated batch processing after all data was ready and considered locked." This indicates the algorithm processed the data without human intervention to influence its initial measurements. Manual adjustments are available in the device, but the reported study focuses on the automated output.
  • LVivo Bladder: The algorithm was applied "by an automated batch processing on all pairs of trans and long views," also indicating standalone performance for the reported results. The device allows for manual caliper adjustments, but the validation appears to have used the automated result.

7. The Type of Ground Truth Used

• LVivoRV: Expert consensus/manual measurements from qualified medical professionals (sonographers and echo cardiologists) based on conventional echocardiography methods (2D manual measurements, M-Mode, VVI).
• LVivo Bladder: Expert manual measurements by an expert sonographer, considered the routinely used method for bladder volume measurement.

8. The Sample Size for the Training Set

• The document does not explicitly state the sample size used for the training set for either LVivoRV or LVivo Bladder. It only describes the algorithms and their application to the test sets.

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

• The document does not provide details on how the ground truth for the training set was established. It describes the technology for LVivoRV as combining "image processing and Deep Learning Neural Network (NN)" and for LVivo Bladder as using "a combination of machine learning and active contour," implying the use of training data, but no specifics are given regarding its ground truth establishment.

Ask a specific question about this device