Search Results

Clarius Bladder AI is intended for semi-automatic non-invasive measurements of bladder volume on ultrasound data acquired by the Clarius Ultrasound Scanner (i.e., curvilinear and phased array scanners). The user shall be a healthcare professional trained and qualified in ultrasound. The ultimate responsibility of ascertaining the measurements based on standard practices and clinical judgment. Clarius Bladder AI is indicated for use in adult patients only.

Clarius Bladder AI is a radiological (ultrasound) image processing software application which implements artificial intelligence (Al), utilizing non-adaptive machine learning algorithms, and is incorporated into the Clarius App software for use as part of the complete Clarius Ultrasound Scanner system product offering in bladder ultrasound imaging applications. Clarius Bladder Al is intended for use by trained healthcare practitioners for non-invasive measurements of bladder volume on ultrasound data acquired by the Clarius Ultrasound Scanner system (i.e., curvilinear and phased array scanners) using an artificial intelligence (AI) image segmentation algorithm.

During the ultrasound imaging procedure, the anatomical site (bladder) is selected through a preset software selection (i.e., bladder) within the Clarius App in which Clarius Bladder Al will engage to segment the bladder and place calipers for calculation of bladder volume.

Clarius Bladder Al operates by performing the following automations:

• . Automatic detection and measurement of bladder depth
• . Automatic detection and measurement of bladder width
• . Automatic detection and measurement of bladder height
• . Automatic detection of the corresponding image view (sagittal vs. transverse)

Clarius Bladder Al operates by performing automatic measurements of bladder height, width, and length, and calculates bladder volume. The user has the option to manually adjust the measurements made by Clarius Bladder Al by moving the caliper crosshairs. Clarius Bladder Al does not perform any functions that could not be accomplished manually by a trained and qualified user. Clarius Bladder Al is intended for use in B-Mode only.

Clarius Bladder AI is an assistive tool intended to inform clinical management and is not intended to replace clinical decision-making. The clinician retains the ultimate responsibility of ascertaining the measurements based on standard practices and clinical judgment. Clarius Bladder Al is indicated for use in adult patients only.

Clarius Bladder AI is incorporated into the Clarius App software, which is compatible with iOS and Android operating systems two versions prior to the latest iOS or Android stable release build and is intended for use with the following Clarius Ultrasound Scanner system transducers (previously 510(k)-cleared in K213436). Clarius Bladder Al is not a stand-alone software device.

Here's a breakdown of the acceptance criteria and the study details for the Clarius Bladder AI device, based on the provided FDA 510(k) summary:

1. Table of Acceptance Criteria and Reported Device Performance

The core acceptance criterion for Clarius Bladder AI's automated measurements was non-inferiority to manual measurements performed by qualified experts, with an equivalence margin of 25% for the mean difference between percentage differences of bladder volume measurements.

Prospective Study: p-value of 1.36e-14 (confirming non-inferiority). Mean difference between percent differences of clinical expert mean and Bladder AI mean was -0.0228 (95% CI -0.074, 0.028). |
| Agreement with Experts: Strong agreement between Clarius Bladder AI measurements and the mean of expert clinicians' measurements, and with individual expert measurements. | Both retrospective and prospective studies reported strong agreement between Clarius Bladder AI and expert measurements, as well as high inter-rater reliability (Intraclass Correlation Coefficients for inter-rater reliability were calculated and found to be strong). Average Dice scores and Jaccard index were also calculated, indicating good segmentation agreement. |
| Clinical Usability: Performs as intended in a representative user environment, meets product requirements, is clinically usable, and meets user needs for semi-automated bladder volume measurements. | Clinical validation study results showed consistent results among all users, meeting pre-defined acceptance criteria, demonstrating that Clarius Bladder AI performs as intended and meets user needs. Users were able to activate, image, perform live segmentation, automatic measurements, manual adjustments, and save measurements. |

2. Sample Size and Data Provenance for Test Set

Retrospective Study:

• Sample Size: 66 subjects (10 female, 38 male, gender of remaining unknown)
• Data Provenance: Anonymized multi-center database of images from predominantly the United States. Institutions included in the model training and tuning datasets were excluded from this study. Retrospective.

Prospective Study:

• Sample Size: 58 subjects (40 female, 18 male)
• Data Provenance: Conducted at a healthcare institution in the United States. Images were obtained prospectively.

3. Number of Experts and Qualifications for Ground Truth

• Number of Experts: 3 reviewers (referred to as "clinical truthers" or "clinical experts") for both the retrospective and prospective studies.
• Qualifications of Experts: Described as "qualified experts with relevant (i.e., bladder) ultrasound experience."
  • For the retrospective study: "qualified experts with relevant (i.e., bladder) ultrasound experience."
  • For the prospective study: "qualified experts with clinical experience in bladder ultrasound."

4. Adjudication Method for the Test Set

The ground truth for bladder volume in both retrospective and prospective studies was established as the mean bladder volume measurement among the three clinical experts. Each reviewer was blinded to the Clarius Bladder AI output and the other reviewers' annotations.

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

The provided information does not explicitly describe a traditional MRMC comparative effectiveness study designed to measure the effect size of human readers improving with AI vs. without AI assistance.

Instead, the studies focused on demonstrating the non-inferiority of the AI device's standalone measurements compared to the mean of multiple human expert measurements. While comparisons were made between reviewer pairs (inter-rater reliability), and between the AI output and individual/mean expert measurements, the studies did not seem to directly evaluate human performance with the AI assistance versus human performance without it in a controlled MRMC setting to quantify a "human improvement" effect size.

6. Standalone Performance Study (Algorithm Only)

Yes, a standalone (algorithm only) performance study was done. The core of both the retrospective and prospective verification studies was to evaluate the Clarius Bladder AI's automated measurements directly against expert manual measurements, demonstrating its performance without human intervention (other than initial image acquisition and potential later manual adjustment by the user, which was a separate feature). The non-inferiority claims are based on this standalone performance.

7. Type of Ground Truth Used

The ground truth used was expert consensus, specifically defined as the mean bladder volume measurement among three clinical experts.

8. Sample Size for the Training Set

• Training Dataset: 1352 subjects (353 female, 999 male).
  • Note: This also includes a validation (tuning) dataset which was 10% of the training data.

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

The deep neural network (DNN) model was trained using the raw training dataset. The summary states that the test data (which was independent) was "labelled by experts." While it doesn't explicitly detail the ground truth establishment for the training set itself, it can be inferred that similar expert labeling or a robust annotation process would have been used to generate the ground truth for the images used in training. The summary highlights that the validation (tuning) data was independent, and the test data was "labelled by experts," suggesting expert annotation for ground truth across relevant datasets. The overall context points to expert-derived ground truth for model development.

Ask a specific question about this device