(271 days)
DeepCT is a notification-only, parallel workflow tool for use by hospital networks and trained clinicians to identify and communicate images of specific patients to a specialist, independent of standard of care workflow. DeepCT uses an artificial intelligence algorithm to analyze images for findings suggestive of a pre-specified clinical condition and to notify an appropriate medical specialist of these findings in parallel to standard of care image interpretation. Identification of suspected findings is not for diagnostic use beyond notification. Specifically, the device analyzes non-contrast CT images of the brain acquired in the acute setting and sends notifications to a specialist that a suspected ICH (intracranial hemorrhage) has been identified and recommends review of those images. Notified clinicians are responsible for viewing non-contrast CT images of the brain on a diagnostic viewer and engaging in appropriate patient evaluation and relevant discussion with a treating specialist before making care-related decisions or requests. DeepCT is limited to analysis of imaging data and should not be used in-lieu of full patient evaluation or relied upon to make or confirm diagnosis.
This software is used to analyze the head computed tomography image of a patient suspected of having intracranial hemorrhage and/or hematoma (hereinafter referred to as "ICH"). Provide a "present" situation (with ICH) notification, send a text message to the user.
DeepCT (Ver. 4.1.4) is a software-only device that uses two components: (1) Image Forwarding Software and (2) Image Processing and Analysis Server.
(1) The Image Forwarding Software is configured by the hospital to be used on a computer and is responsible for transmitting a copy of DICOM files from the local through a secured channel to the Image Processing and Analysis Server.
When the Image Forwarding Software receives the interpretation result from the Image Processing and Analysis Server, it shows the result on the screen. If there is a suggestive of ICH, the Image Forwarding Software sends a notification to the specialist identifying the study of interest. While the software informs the notification process, no other diagnostic information is generated from the software or available to the user beyond the notification.
(2) The Image Processing and Analysis Server is responsible for receiving, assembling, processing, analyzing and storing DICOM images. This component includes the algorithm that is responsible for identifying and quantifying image characteristics that are consistent with a ICH and transmit the result back to the Image Forwarding Software.
Here's a breakdown of the acceptance criteria and the study details for the DeepCT device, based on the provided text:
1. Acceptance Criteria and Reported Device Performance
| Acceptance Criteria | Reported Device Performance |
|---|---|
| Sensitivity $\ge$ 80% | 93.8% (95% CI: 88.3%-96.8%) |
| Specificity $\ge$ 80% | 92.3% (95% CI: 86.4%-95.7%) |
| Processing Time | 30.6 seconds (95% CI: 25.8-35.4 seconds), which is lower than the processing time reported by the Aidoc BriefCase device (the predicate device, though exact Aidoc time not provided). |
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size: 260 cases
- Data Provenance: Retrospective, multicenter, multinational.
- Countries: 5 clinical sites (2 US and 3 OUS - Outside US). Specific countries are not mentioned beyond "US" and "OUS".
- Distribution: 130 cases from US sites and 130 cases from OUS sites.
- Case Balance: Approximately an equal number of positive (images with ICH) and negative (images without ICH) cases.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications
The document does not specify the number of experts used to establish ground truth for the test set or their qualifications. It only states that the study evaluated "the software's performance in identifying non-contrast CT head images containing ICH findings," implying an established ground truth, but details are absent.
4. Adjudication Method for the Test Set
The document does not specify the adjudication method used for the test set.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
No. The document describes a standalone study evaluating the algorithm's performance against a pre-established ground truth. It does not mention a comparative effectiveness study involving human readers with and without AI assistance.
6. Standalone Performance Study
Yes. The study described in the "Performance Testing" section is a standalone study of the algorithm's performance. It evaluates DeepCT's sensitivity, specificity, and processing time in identifying ICH without human-in-the-loop performance measurement.
7. Type of Ground Truth Used
The document implies a ground truth based on the presence or absence of "ICH findings" in the images, but it does not explicitly state the method used to establish this ground truth (e.g., expert consensus, pathology, outcomes data).
8. Sample Size for the Training Set
Radiology records were collected from 21,603 patients who underwent head CT scans between 2007 and 2017. This dataset was used for DeepCT development and deployment. It is implied this was the training set, or at least a significant portion of it.
9. How the Ground Truth for the Training Set Was Established
The document states: "The Tri-Service General Hospital Institutional Review Board, Kaohsiung Veterans General Hospital Institutional Review Board and National Taiwan University Hospital Research Ethics Committee all approved and consented the use of the retrospective image data for DeepCT development and deployment without relevant ethical concern."
While Institutional Review Board (IRB) approval is mentioned for the use of the data, the document does not explicitly describe how the ground truth labels (i.e., presence or absence of ICH) were established for this large training dataset. It only refers to "radiology records" and the "retrospective image data." It's highly probable that these labels were derived from radiologists' interpretations in the original radiology reports, but this is not definitively stated.
§ 892.2080 Radiological computer aided triage and notification software.
(a)
Identification. Radiological computer aided triage and notification software is an image processing prescription device intended to aid in prioritization and triage of radiological medical images. The device notifies a designated list of clinicians of the availability of time sensitive radiological medical images for review based on computer aided image analysis of those images performed by the device. The device does not mark, highlight, or direct users' attention to a specific location in the original image. The device does not remove cases from a reading queue. The device operates in parallel with the standard of care, which remains the default option for all cases.(b)
Classification. Class II (special controls). The special controls for this device are:(1) Design verification and validation must include:
(i) A detailed description of the notification and triage algorithms and all underlying image analysis algorithms including, but not limited to, a detailed description of the algorithm inputs and outputs, each major component or block, how the algorithm affects or relates to clinical practice or patient care, and any algorithm limitations.
(ii) A detailed description of pre-specified performance testing protocols and dataset(s) used to assess whether the device will provide effective triage (
e.g., improved time to review of prioritized images for pre-specified clinicians).(iii) Results from performance testing that demonstrate that the device will provide effective triage. The performance assessment must be based on an appropriate measure to estimate the clinical effectiveness. The test dataset must contain sufficient numbers of cases from important cohorts (
e.g., subsets defined by clinically relevant confounders, effect modifiers, associated diseases, and subsets defined by image acquisition characteristics) such that the performance estimates and confidence intervals for these individual subsets can be characterized with the device for the intended use population and imaging equipment.(iv) Stand-alone performance testing protocols and results of the device.
(v) Appropriate software documentation (
e.g., device hazard analysis; software requirements specification document; software design specification document; traceability analysis; description of verification and validation activities including system level test protocol, pass/fail criteria, and results).(2) Labeling must include the following:
(i) A detailed description of the patient population for which the device is indicated for use;
(ii) A detailed description of the intended user and user training that addresses appropriate use protocols for the device;
(iii) Discussion of warnings, precautions, and limitations must include situations in which the device may fail or may not operate at its expected performance level (
e.g., poor image quality for certain subpopulations), as applicable;(iv) A detailed description of compatible imaging hardware, imaging protocols, and requirements for input images;
(v) Device operating instructions; and
(vi) A detailed summary of the performance testing, including: test methods, dataset characteristics, triage effectiveness (
e.g., improved time to review of prioritized images for pre-specified clinicians), diagnostic accuracy of algorithms informing triage decision, and results with associated statistical uncertainty (e.g., confidence intervals), including a summary of subanalyses on case distributions stratified by relevant confounders, such as lesion and organ characteristics, disease stages, and imaging equipment.