Search Results

AI-Rad Companion (Pulmonary) is image processing software that provides quantitative and qualitative analysis from previously acquired Computed Tomography DICOM images to support radiologists and physicians from specialty care and general practice in the evaluation and assessment of disease of the lungs.

It provides the following functionality:

• Segmentation and measurements of complete lung and lung lobes
• · Identification of areas with lower Hounsfield values in comparison to a predefined threshold for complete lung and lung lobes
• · Providing an interface to external Medical Device syngo.CT Lung CAD
• · Segmentation and measurements of solid and sub-solid lung nodules
• Dedication of found lung nodules to corresponding lung lobe
• Correlation of segmented lung nodules of current scan with known priors and quantitative assessment of changes of the correlated data.
• Identification of areas with elevated Hounsfield values, where areas with elevated versus high opacities are distinquished.

The software has been validated for data from Siemens Healthineers (filtered backprojection and iterative reconstruction), GE Healthcare (filtered backprojection reconstruction), and Philips (filtered backprojection reconstruction).

Only DICOM images of adult patients are considered to be valid input.

The subject device AI-Rad Companion (Pulmonary) is an image processing software that utilizes machine learning and deep learning algorithms to provide quantitative and qualitative analysis from previously acquired Computed Tomography DICOM images to support qualified clinicians in the evaluation and assessment of disease of the thorax. AI-Rad Companion (Pulmonary) builds on platform functionality provided by the AI-Rad Companion Engine and cloud/edge functionality provided by the Siemens Healthineers teamplay digital platform. AI-Rad Companion (Pulmonary) is an adjunct tool and does not replace the role of a qualified medical professional. AI-Rad Companion (Pulmonary) is also not designed to detect the presence of radiographic findings other than the prespecified list. Qualified medical professionals should review original images for all suspected pathologies.

AI-Rad Companion (Pulmonary) offers:

• Segmentation of lungs, ●
• Segmentation of lung lobes.
• Parenchyma evaluation, ●
• Parenchyma ranges,
• Pulmonary density,
• Visualization of segmentation and parenchyma results,
• Interface to LungCAD,
• Lesion segmentation, ●
• Visualization of lesion segmentation results, ●
• Lesion follow-up

AI-Rad Companion (Pulmonary) requires images of patients of 22 years and older.

AI-Rad Companion (Pulmonary) SW version VA40 is an enhancement to the previously cleared device AI-Rad Companion (Pulmonary) (K213713) that utilizes machine and deep learning algorithms to provide quantitative and qualitative analysis to computed tomography DICOM images to support qualified clinicians in the evaluation and assessment of disease of the thorax.

As an update to the previously cleared device, the following modifications have been made:

• Sub-solid Lung Nodule Segmentation ●

This feature provides the ability to segment and measure all subtypes of lesions including solid and sub-solid lesions.

• . Modified Indications for Use Statement The indications for use statement was updated to include descriptive text for sub-solid lung nodule addition.
• Updated Subject Device Claims List The claims list was updated to reflect the new device functionality
• . Updated Limitations for Use Additional limitations for use has been added to the subject device.

Here's a breakdown of the acceptance criteria and the study proving the device's performance, based on the provided text:

1. Table of Acceptance Criteria and Reported Device Performance

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

• Sample Size for Test Set: 273 subjects from the United States and 254 subjects from Germany, for a total of 527 subjects.
• Data Provenance: The data originated from the United States (69% of cases) and Germany (31% of cases). The data was retrospective, as it refers to "previously acquired Computed Tomography DICOM images."
• Imaging Vendors: The test data included images from Canon/Toshiba (18%), GE (35%), Philips (15%), and Siemens (32%).

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

• Number of Experts: Two board-certified radiologists, with a third radiologist for adjudication.
• Qualifications:
  • Radiologist 1: 10 years of experience (board-certified)
  • Radiologist 2: 7 years of experience (board-certified)
  • Adjudicating Radiologist 3: 9 years of experience

4. Adjudication Method for the Test Set

• Method: 2+1 (Two experts independently established ground truth, and in case of disagreement, a third expert served as an adjudicator.)

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done

• Was it done?: No, a traditional MRMC comparative effectiveness study involving human readers was not performed in the context of this specific submission. The study focuses on the standalone performance of the AI algorithm in comparison to the predicate device's performance, particularly for the new sub-solid nodule segmentation feature. The device is described as an "adjunct tool," but the presented study validates the algorithm's performance against expert consensus, not against human readers with and without AI.

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

• Was it done?: Yes, the performance testing described directly evaluates the AI-Rad Companion (Pulmonary) lesion segmentation algorithm's accuracy (measured by DICE score, bias, and RMSE) against established ground truth. This is a standalone performance evaluation of the algorithm.

7. The Type of Ground Truth Used

• Type: Expert Consensus. The ground truth annotations for the test data were established independently by two board-certified radiologists, with a third radiologist serving as an adjudicator in cases of disagreement.

8. The Sample Size for the Training Set

• The sample size for the training set is not explicitly stated in the provided document. However, it is mentioned that "None of the clinical sites providing the test data provided data for training of any of the algorithms. Therefore there is a clear independence on site level between training and test data." This indicates that a distinct training set (or sets) was used.

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

• The document does not explicitly state how the ground truth for the training set was established. It only emphasizes the independence of the training and test data sites.

Ask a specific question about this device

AI-Rad Companion (Pulmonary) is image processing software that provides quantitative and qualitative analysis from previously acquired Computed Tomography DICOM images to support radiologists and physicians from emergency medicine, specialty care, urgent care, and general practice in the evaluation and assessment of disease of the lungs.

It provides the following functionality:

· Segmentation and measurements of complete lung and lung lobes

• · Identification of areas with lower Hounsfield values in comparison to a predefined threshold for complete lung and lung lobes
• · Providing an interface to external Medical Device syngo.CT Lung CAD
• · Segmentation and measurements of solid lung nodules
• · Dedication of found lung nodules to corresponding lung lobe
• · Correlation of segmented lung nodules of current scan with known priors and quantitative assessment of changes of the correlated data.
• · Identification of areas with elevated Hounsfield values, where areas with elevated versus high opacities are distinguished.

The software has been validated for data from Siemens (filtered backprojection and iterative reconstruction), GE Healthcare (filtered backprojection reconstruction), and Philips (filtered backprojection reconstruction).

Only DICOM images of adult patients are considered to be valid input.

AI-Rad Companion (Pulmonary) SW version VA20 is an enhancement to the previously cleared device AI-Rad Companion (Pulmonary) K183271 that utilizes machine and deep learning algorithms to provide quantitative and qualitative analysis to computed tomography DICOM images to support qualified clinicians in the evaluation and assessment of disease of the thorax.

As an update to the previously cleared device, the following modifications have been made:

• Modified Indications for Use Statement The indications for use statement was updated to include descriptive text for the lung lesion follow feature.
• Updated Subject Device Claims List The claims list was updated to include claim pertaining to the lung lesion follow up feature.
• Lung Lesion Follow-up Assessment of current and prior lesions This feature provides the possibility to compare currently segmented lung lesions with corresponding priors and changes to the correlated data are assessed quantitatively.
• Pulmonary Density Assessment

This feature provides the possibility to segment opacity regions inside the lung using an AI algorithm. AI-Rad Companion (Pulmonary) counts image voxels inside opacity regions and calculates the percentages of these voxels relative to the total number of voxels per lobe. lung and in total. Afterwards, each of the five lung lobes is assigned a score ranging from 0 to 4 based on the percentage of opacity as follows: 0 (0%), 1 (1%-25%), 2 (26%-50%), 3 (51%-75%), or 4 (76%-100%). Then a summation of the five lobe scores (range of possible scores, 0-20) are generated in the device outputs. This functionality is commercially available on the Siemens syngo.CT Extended Functionality (K203699).

• . Bi-Directional Lesion Diameter
  This feature provides an additional measurement derived from the existing segmentation contour of a lung lesion. The existing list of measurements is extended with the maximum orthogonal diameter in 2D (short axis diameter) which is orthogonal to the lesion's maximum 2D diameter (2D diameter, long axis diameter). This functionality is commercially available on the Siemens syngo.CT Extended Functionality (K203699).

• . Architecture Enhancement for on premise Edge deployment

  • The system supports the existing cloud deployment as well as an on premise "edge" deployment. The system remains hosted in the teamplay digital health platform and remains driven by the AI-Rad Companion Engine. Now the edge deployment implies that the processing of clinical data and the generation of results can be performed onpremises within the customer network. The edge system is fully connected to the cloud for monitoring and maintenance of the system from remote.

The provided document, a 510(k) summary for Siemens Healthcare GmBh's AI-Rad Companion (Pulmonary) SW version VA20, describes the device, its intended use, and the non-clinical tests performed to demonstrate its safety and effectiveness.

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

Acceptance Criteria and Reported Device Performance

The document does not explicitly present a table of acceptance criteria with corresponding performance metrics for all functionalities. However, it does state some performance metrics for one specific feature:

Missing Information: The document does not provide acceptance criteria or performance results for other key functionalities of the AI-Rad Companion (Pulmonary), such as:

• Segmentation and measurements of complete lung and lung lobes.
• Identification of areas with lower Hounsfield values.
• Segmentation and measurements of solid lung nodules.
• Dedication of found lung nodules to corresponding lung lobe.
• Identification of areas with elevated Hounsfield values (Pulmonary Density Assessment).
• Bi-directional lesion diameter measurements.

Study Details:

The provided document describes a non-clinical bench test specifically for the lesion follow-up feature. It explicitly states: "No clinical tests were conducted to test the performance and functionality of the modifications introduced within AI-Rad Companion (Pulmonary)." This implies that the reported performance metrics are from an algorithm-only (standalone) performance evaluation, without human-in-the-loop.

Here's what can be extracted about the study:

1. Sample Size and Data Provenance:

   • Test Set Sample Size: 200 cases were used to evaluate the lesion follow-up feature.
   • Data Provenance: Not explicitly stated. The document mentions validation for data from Siemens, GE Healthcare, and Philips (reconstruction types specified), but it does not specify the country of origin of the data or whether it was retrospective or prospective.

2. Number of Experts and Qualifications:

   • The document does not provide information on the number of experts used to establish ground truth or their specific qualifications for the test set.

3. Adjudication Method:

   • The document does not specify any adjudication method (e.g., 2+1, 3+1, none) for the test set. Since it's a non-clinical bench test of the algorithm's ability to identify lesion pairs, it's possible that a different form of ground truth establishment (e.g., based on established physical measurements or derived from existing clinical reports) was used rather than direct expert consensus on each case.

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

   • No MRMC study was done. The document explicitly states: "No clinical tests were conducted to test the performance and functionality of the modifications introduced within AI-Rad Companion (Pulmonary)." Therefore, there is no effect size reported for human readers improving with AI vs. without AI assistance.

5. Standalone (Algorithm Only) Performance:

   • Yes, a standalone study was done for the lesion follow-up feature. The reported sensitivity and PPV are for the algorithm's performance in identifying lesion pairs.

6. Type of Ground Truth Used:

   • The document does not explicitly state the type of ground truth used for the lesion follow-up test. It mentions "evaluation of 200 cases to identify lesion pairs," which suggests that a definitive ground truth for paired lesions was available for these 200 cases. This could range from expert consensus, to prior established measurements, or structured clinical reports that define lesion pairs.

7. Training Set Sample Size:

   • The document does not specify the sample size used for the training set. It only mentions the use of "machine and deep learning algorithms."

8. How Ground Truth for Training Set Was Established:

   • The document does not describe how the ground truth for the training set was established.

Ask a specific question about this device

AI-Rad Companion (Pulmonary) is image processing software that provides quantitative and qualitative analysis from previously acquired Computed Tomography DICOM images to support radiologists and physicians from emergency medicine, specialty care, urgent care, and general practice in the evaluation and assessment of disease of the lungs. It provides the following functionality:

• Segmentation and measurements of complete lung and lung lobes
• Identification of areas with lower Hounsfield values in comparison to a predefined threshold for complete lung and lung lobes
• Providing an interface to external Medical Device syngo.CT Lung CAD
• Segmentation and measurements of found lung lesions and dedication to corresponding lung lobe.
  The software has been validated for data from Siemens (filtered backprojection and iterative reconstruction), GE Healthcare (filtered backprojection reconstruction), and Philips (filtered backprojection reconstruction).
  Only DICOM images of adult patients are considered to be valid input.

AI-Rad Companion (Pulmonary) is a software only image processing application that supports quantitative and qualitative analysis of previously acquired CT DICOM Images to support radiologists and physicians from emergency medicine, specialty care, and general practice in the evaluation of and assessment of disease of the thorax.

Here is a summary of the acceptance criteria and the study that proves the device meets them, based on the provided FDA document for Siemens AI-Rad Companion (Pulmonary):

1. Table of Acceptance Criteria & Reported Device Performance

The document doesn't explicitly state "acceptance criteria" as clear pass/fail thresholds for each metric. Instead, it describes validated performance results and claims they are "superior" to the predicate device, thereby supporting substantial equivalence. The key performance metrics are for lung lobe segmentation.

• Ranged between 0.95 and 0.98.
• Standard Deviation (SD) 4,500 CT data sets"

• Data Provenance: Retrospective performance study from "multiple clinical sites from within and outside United States."

3. Number of Experts and Qualifications for Ground Truth

• The document does not explicitly state the "number of experts" or their specific "qualifications" beyond mentioning "manually established ground truth." It implies that the ground truth was established by qualified professionals, likely radiologists or trained medical personnel, given the nature of the task (segmentation).

4. Adjudication Method for the Test Set

• The document does not specify an adjudication method (e.g., 2+1, 3+1). It states "manually established ground truth," which typically implies consensus among multiple readers or a single highly experienced reader whose work is considered the gold standard within the study's context.

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

• No evidence of a MRMC study. The document describes a standalone (algorithm only) performance study directly comparing the AI algorithm's output to "manually established ground truth" and claiming superiority over a predicate device's algorithm, not an AI-assisted human reader study. The purpose of this submission is to demonstrate substantial equivalence of the new AI-Rad Companion to existing predicate devices, not improvement in human reader performance.

6. Standalone (Algorithm Only) Performance Study

• Yes, a standalone study was done. The performance metrics (DICE coefficients, surface distance, Hausdorff distance, volume error) were computed by comparing the output of the algorithm to the manually established ground truth. This confirms it was an algorithm-only performance evaluation without human-in-the-loop.

7. Type of Ground Truth Used

• Expert Consensus/Manual Establishment: The ground truth for the test set was "manually established ground truth." This typically refers to annotations or segmentations performed by human experts (e.g., radiologists) and potentially reviewed for consensus.

8. Sample Size for the Training Set

• The document mentions a "Training cohort: size and properties of data used for training" as a structural element of their analysis but does not provide the specific sample size for the training set.

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

• The document mentions "Description of ground truth / annotations generation" as a structural element for their algorithm analysis but does not detail how the ground truth for the training set was established. It can be inferred that it was likely generated through expert annotations, potentially similar to the test set, but specific methods (e.g., single expert, multiple experts, consensus, specific tools) are not described.

Ask a specific question about this device