Search Results

The CADDIE computer-assisted detection device is intended to assist the gastroenterologist in detecting suspected colorectal polyps only. The gastroenterologist is responsible for reviewing CADDIE suspected polyp areas and confirming the presence or absence of a polyp based on their own medical judgment.

CADDIE is not intended to replace a full patient evaluation, nor is it intended to be relied upon to make a primary interpretation of endoscopic procedures, medical diagnosis, or recommendations of treatment/course of action for patients. The CADDIE computer-assisted detection device is limited for use with standard white-light endoscopy imaging only.

CADDIE is cloud based artificial intelligence medical device software. CADDIE interfaces with the video feed generated by an endoscopic video processor during a colonoscopy procedure

The software is intended to be used by trained and qualified healthcare professionals as an accompaniment to video endoscopy for the purpose of drawing attention to regions with visual characteristics consistent with colonic mucosal lesions (such as polyps and adenomas).

CADDIE analyses the data from the endoscopic video processor in real-time and provides information to aid the endoscopist in detecting suspected colorectal polyps, if they are in the field of view of the endoscope.

The areas highlighted by CADDIE are not to be interpreted as definite polyps or adenomas. The responsibility to make a decision as to whether or not a highlighted region contains a polyp or is an adenoma lies with the user. The endoscopist is responsible for reviewing CADDIE suspected polyp areas and confirming the presence or absence of a polyp and its classification based on their own medical judgement.

Here's a detailed breakdown of the acceptance criteria and the study that proves the device meets them, based on the provided FDA 510(k) Clearance Letter for CADDIE K252586:

1. Table of Acceptance Criteria and Reported Device Performance

Note: The FDA letter explicitly states that for all changes (including the Cecum AI update), "design and development activities were performed in compliance with Odin's Quality Management System and FDA Design Control requirements (21 CFR 820.30), using the same methods and acceptance criteria described in the cleared submission (K240044)." However, the provided document for K252586 does not explicitly list the quantitative acceptance criteria (e.g., minimum TPR, maximum FPR) from the original K240044 submission for these specific metrics of the Cecum AI feature. It only reports the achieved performance. The implicit acceptance is that the device's performance is "satisfactory and do not raise any additional questions on the safety and effectiveness" as compared to the predicate.

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

• Sample Size: The test set for the Cecum AI Convenience Feature Standalone Bench-testing Dataset comprised 5733 frames. This was broken down into:
  • 838 Positive Frames (containing Cecal structures)
  • 461 frames with Appendiceal Orifice (AO)
  • 418 frames with Ileocecal Valve (ICV)
  • 4016 Negative Frames (without Cecal structures)
• Data Provenance: The document states the dataset consists of "photo-documented frames from a standard colonoscopy procedure" and mentions "historical control (known cecal structure status per frame)." It does not explicitly state the country of origin or if it's retrospective or prospective. However, given it's "historical control" and "recorded colonoscopy frames," it is highly likely to be retrospective data.

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

• Number of Experts: A "team of trained clinical annotators" was used. The exact number is not specified.
• Qualifications: They are described as "trained clinical annotators." Specific professional qualifications (e.g., radiologist, gastroenterologist) or years of experience are not provided.

4. Adjudication Method for the Test Set

• The document states, "Annotation was performed on a per-frame basis, where a team of trained clinical annotators labelled cecal structures with a bounding box. These annotations were used as ground truth reference standards."
• It does not explicitly state an adjudication method (like 2+1 or 3+1). It implies that the "team of trained clinical annotators" collectively established the ground truth, but the process for resolving disagreements or reaching consensus is not detailed.

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

• No, an MRMC comparative effectiveness study was not done for the Cecum AI convenience feature.
• The primary submission (K252586) is a Special 510(k) addressing an update to a "convenience feature" (Cecum AI model algorithm). It explicitly states that "The baseline Clinical Performance Evaluation was conducted and reviewed in K240044 and is still applicable to the versions of the device that are the subject of this submission." This suggests that any MRMC studies related to the primary polyp detection functionality would have been part of the original K240044 clearance. The current submission focuses only on the AI model change for cecal landmark detection.

6. Standalone Performance Study (Algorithm Only Without Human-in-the-Loop)

• Yes, a standalone performance study was done for the Cecum AI Convenience Feature.
• The document states, "Standalone performance testing was performed to assess the ability of the Cecum AI Convenience Feature to discriminate between normal mucosa and cecal landmarks... A set of recorded colonoscopy frames were analyzed by the Cecum AI Convenience Feature and the results were compared to the historical control (known cecal structure status per frame)." The reported metrics (accuracy, TPR, FPR, AUC) are all standalone performance metrics.

7. Type of Ground Truth Used

• The ground truth used for the Cecum AI convenience feature was expert consensus/annotation. Specifically, "Annotation was performed on a per-frame basis, where a team of trained clinical annotators labelled cecal structures with a bounding box. These annotations were used as ground truth reference standards."

8. Sample Size for the Training Set

• The document states that the "Non-clinical performance testing was performed on the standalone bench-testing dataset, which is separate to the development datasets."
• However, it does not provide the sample size for the training set used to develop the Cecum AI model.

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

• The document does not explicitly describe how the ground truth for the training set was established. It only details the ground truth for the test set. It is reasonable to infer that a similar process of expert annotation would have been used for the training data, but this is not stated.

Ask a specific question about this device