Search Results

CT Cardiomegaly is a command line software application intended to be run on its own or as part of another medical device to automatically calculate linear and area based cardiothoracic ratio (CTR) from a CT image containing the heart.

CT Cardiomegaly is designed to measure the maximal transverse diameter of heart and maximal inner transverse diameter of thoracic cavity and calculate the CTR from an axial CT slice containing the heart using a non-adaptive machine learning algorithm.

Intended users of the software are aimed to the physicians or other licensed practitioners in the healthcare institutions, such as clinics, hospitals, healthcare facilities, residential care facilities and long-term care services.

The system is suitable for adults and transitional adolescents (18 to 21 years old but treated as an adult).

Its results are not intended to be used on a stand-alone basis for clinical-decision making or otherwise preclude clinical assessment of any disease.

The CT Cardiomegaly application is a software only (SaMD) medical device which includes automated algorithms and non-adaptive machine learning to analyze chest computed tomography (CT) images. CT Cardiomegaly is a command line software intended to be run on its own or as part of another medical device.

CT Cardiomegaly scans the existing images for the presence of specific anatomical structures (i.e., heart and chest cavity). If the target structures are present, the software segments and performs automated measurements which may be helpful in assisting with the detection of certain diseases and/or conditions (i.e., cardiomegaly). The quantitative information is provided to the clinician in the form of a summary report. The clinician (using all available information) then decides if further diagnostic work-up, analysis or measurements are indicated.

The results from CT Cardiomegaly are not intended to be used as the primary input for clinicaldecision making or diagnosis.

A reference standard was created to train the machine learning algorithms in CT Cardiomegaly. Data for training and testing were sampled from this database based on exclusion criteria. Data was also selected to ensure no site was shared by training and testing cohorts. The training dataset included 1600 unique patient cases from 267 unique sites across 20 U.S. states.

Here's a breakdown of the acceptance criteria and the study that proves the device meets them, based on the provided text:

Acceptance Criteria and Device Performance

Study Details

1. A table of acceptance criteria and the reported device performance:
See table above. The acceptance criteria for the primary endpoints are defined by "Mean Difference Target" and for the secondary endpoint (Key Slice Position) by "Mean Difference Target". For segmentation, "Observed Dice [95% CI]" serves as the performance metric.

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

• Test Set Sample Size: 275 unique patient cases.
• Data Provenance: Data was sampled from a database curated by the University of Alabama at Birmingham. The test set included cases from sites across 12 U.S. states. The dataset included various patient demographics (age, sex, race, ethnicity), confounding conditions, and imaging characteristics (protocol, slice thickness, scanner manufacturer, presence or absence of contrast, kVp), making it representative of diverse clinical scenarios.

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

• Number of Experts: Three.
• Qualifications: All three experts were board-certified radiologists.

4. Adjudication method for the test set:

• The text states that each radiologist "provided a segmentation for the heart and the inner chest independently from the other radiologists." It also mentions that "Measurement among the three radiologists was very high for all measured values." This implies that while initial segmentations were independent, a consensus or average of these high-agreement measurements was used as the reference standard, rather than a specific 2+1 or 3+1 adjudication rule.

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, an MRMC comparative effectiveness study that assesses human reader improvement with AI assistance versus without AI assistance was not explicitly described. The study focused on the standalone performance of the device against a human-derived reference standard.

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

• Yes, a standalone performance assessment was conducted. The study validated the accuracy of the CT Cardiomegaly's measurements (linear-based CTR and area-based CTR), key heart slice detection, and segmentation accuracy ("algorithm only") against the measurements provided by human readers (the reference standard).

7. The type of ground truth used:

• Expert Consensus: The ground truth was established by segmentations and measurements provided by three board-certified radiologists, whose measurements showed "very high" agreement. This falls under expert consensus.

8. The sample size for the training set:

• 1600 unique patient cases.

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

• A reference standard was created to train the machine learning algorithms in CT Cardiomegaly. While the document doesn't explicitly detail the method for establishing the training set ground truth, it's reasonable to infer that a similar process involving expert radiologists and their segmentations/measurements would have been used, perhaps with subsequent review or consensus, given the explicit mention of a reference standard being created for training. The document states, "Data for training and testing were sampled from this database based on exclusion criteria. Data was also selected to ensure no site was shared by training and testing cohorts."

Ask a specific question about this device