Search Results

The ME-APDS (Magentig Eye's Automatic Polyp Detection System) is intended to be used by endoscopists as an adjunct to the common video colonoscopy procedure (screening and surveillance), aiming to assist in identifying lesions during colonoscopy procedure by highlighting regions with visual characteristics consistent with different types of mucosal abnormalities that appear in the colonoscopy video during the procedure. Highlighted regions can be independently assessed by the endoscopist and appropriate action taken according to standard clinical practice.

The ME-APDS is trained to process video images which may contain regions consistent with polyps.

The ME-APDS is limited for use with standard white-light endoscopy imaging only.

The ME-APDS is intended to be used as an adjunct to endoscopy procedures and is not intended to replace histopathological sampling as means of diagnosis.

Device Description

The ME-APDS (Magentig Eve's Automatic Polvo Detection System) is intended to be used as an adjunct to the common video colonoscopy procedure. The system application aims to assist the endoscopist in identifying lesions, such as polyps, during the colonoscopy procedures in real time. The device is not intended to be used for diagnosis or characterization of lesions, and does not replace clinical decision making.

The system acquires the digital video output signal from the local endoscopy camera and processes the video frames. It runs deep machine learning and additional supporting algorithms in real time on the video frames in order to detect and identify regions having characteristics consistent with different tvpes of mucosal abnormalities such as polyps. The output video with the detected lesions is presented on a separate touchscreen, supplied as part of the ME-APDS, highlighting the suspicious areas on the original video. The output of the system can also be presented on additional monitors in the procedure room using the 1x4 HDMI Splitter supplied with the system. The user can also take snapshots of the videos, with and without the highlighting of the suspicious areas, record videos and view in full screen mode.

AI/ML Overview

Acceptance Criteria and Study Details for ME-APDS™

1. Table of Acceptance Criteria and Reported Device Performance:

Metric	Acceptance Criteria (Stated or Implied)	Reported Device Performance (ME-APDS™)
Standalone Performance:
Polyp-wise Recall (Polyps with Histology)	Not explicitly stated, but high recall across consecutive frames is implied for adequate aid.	PRecall1: 100.0%
PRecall3: 99.6%
PRecall5: 99.6%
PRecall7: 99.6%
Polyp-wise Recall (Entire Testing Dataset)	Not explicitly stated, but high recall across consecutive frames is implied for adequate aid.	PRecall1: 98.2%
PRecall3: 94.2%
PRecall5: 91.5%
PRecall7: 90.0%
False Positives Per Full Video (FPPF)	FPPF threshold of 0.0328 (normalized to 15 minutes)	Met the FPPF threshold of 0.0328
Marker Annotation Latency	Not explicitly stated, but real-time performance is a key feature.	Median: 0.166 sec (5 frames)
Average: 0.85 sec
Robustness (IoU threshold variation)	Robust performance with varying IoU thresholds up to 0.2.	Changing IoU from 0.01 to 0.1 and 0.2 "slightly influenced only the framewise recall, and did not influence the other results supporting the robustness of the testing."
Clinical Performance (Comparative Effectiveness):
Adenomas Per Colonoscopy (APC)	MEAC APC expected to be >1.05 compared to CC APC (lower limit of 95% CI of MEAC/CC ratio).	MEAC APC: 0.70
CC APC: 0.51 (Implied ratio: 0.70/0.51 ≈ 1.37, which is > 1.05)
Adenomas Per Extraction (APE)	APE of MEAC expected to be non-inferior to APE of CC (lower limit of 95% CI of difference between APE of MEAC and CC expected to be above -0.20).	MEAC APE: 0.31
CC APE: 0.27 (Difference: 0.04. The document states "APE proved non-inferior to that of CC," indicating the criterion was met. Actual CI not explicitly given in table for overall APE difference but for subgroups).

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

Standalone Performance Testing:
- Sample Size: 172 unique full colonoscopy videos, containing 449 polyps (16 videos contained no polyps).
- Data Provenance: Not explicitly stated for each video, but the videos covered various demographic factors (subject sex, age, race). Given that the clinical study data was collected from "10 medical centers in Europe, the United States and Israel," it is highly probable that the standalone testing data also originates from a similar diverse geographical pool. The context suggests it is likely retrospective video data collected from past procedures.
Clinical Testing (Comparative Effectiveness Study):
- Sample Size: 950 patients enrolled (916 patients for baseline demographics). The treatment arms were:
  - CC (Conventional Colonoscopy): 398 patients
  - CC-MEAC (CC followed by MEAC): 69 patients
  - MEAC (ME-APDS-assisted Colonoscopy): 385 patients
  - MEAC-CC (MEAC followed by CC): 64 patients
- Data Provenance: A randomized, two-arm, multi-center, controlled study conducted at 10 medical centers in Europe, the United States, and Israel. This is a prospective study.

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

Standalone Performance Testing: The document does not explicitly state the number of experts or their qualifications for establishing the ground truth specifically for the standalone test set videos. It mentions "Polyps evaluated varied by subject sex... 263 polyps had histology findings..." which implies that histology was used for ground truth. The polyps being "reported in the procedure report" suggests the involvement of clinical experts (endoscopists) at the time of procedure.
Clinical Testing (Comparative Effectiveness Study): The ground truth for polyps (number of adenomas, extractions) was established by the endoscopists performing the colonoscopies within each arm of the study. These would be qualified medical professionals (endoscopists) at the participating clinical centers. The document does not specify their exact years of experience or the number of individual experts beyond the "10 medical centers." The use of "histology" for confirmation of adenomas is also mentioned implicitly in the APE definition, meaning pathology experts contribute to the final ground truth.

4. Adjudication Method for the Test Set:

Standalone Performance Testing: Not explicitly stated. The mention of "polyps verified by histology" implies that a pathologist's report served as the ultimate ground truth. It does not describe an adjudication process between multiple readers of the videos for annotation or ground truth establishment.
Clinical Testing (Comparative Effectiveness Study): No specific adjudication method across multiple independent experts is described for determining the ultimate ground truth in the clinical study. The number of adenomas and extractions were recorded during the colonoscopy procedures, with histology confirming the nature of extracted polyps. The "tandem" design (CC followed by MEAC or MEAC followed by CC) in a subset of patients implicitly allows for a comparison of findings within the same patient, acting as an internal check.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, and effect size of human readers improvement with AI vs without AI assistance:

Yes, a comparative effectiveness study was done.
Study Design: A randomized, two-arm, multi-center, controlled study comparing conventional colonoscopy (CC) with ME-APDS-assisted colonoscopy (MEAC). This is a clinical trial, not a typical MRMC study where multiple readers interpret cases for diagnostic accuracy. However, it still assesses the effectiveness of human readers with AI assistance versus without.
Effect Size:
- Adenomas Per Colonoscopy (APC): MEAC APC was 37% higher (relative increase) than CC APC.
  - MEAC APC: 0.70
  - CC APC: 0.51
  - Absolute difference: 0.70 - 0.51 = 0.19 adenomas per colonoscopy.
- The study found "a mean 0.20 increment between arms for each analyzed subgroup" for APC.
- MEAC was "more effective than CC in detecting ≤5mm polyps and in detecting >6-9 mm polyps, sessile and flat polyps and adenomas in the proximal colon."
- "more sessile serrated adenomas (SSAs) were identified in MEACs as compared to CCs, which resulted in also a higher sessile serrated detection rate (SDR)."

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

Yes, a standalone performance testing was done. This is detailed under "Standalone Performance Testing" on page 5.

7. The type of ground truth used:

Standalone Performance Testing: The ground truth was primarily based on histology findings for 263 polyps and implicitly based on the endoscopist's procedure report for the remaining polyps identified in the videos.
Clinical Testing (Comparative Effectiveness Study): The ground truth was established by recorded adenoma detections and extractions by endoscopists during the procedures, with the definitive diagnosis of adenomas confirmed by histopathology results.

8. The sample size for the training set:

The document does not explicitly state the sample size (number of videos or polyps) used for the ME-APDS training set. It only mentions that the system "runs deep machine learning and additional supporting algorithms" and "is trained to process video images."

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. However, given the nature of the device and the testing methodologies, it is highly probable that the training data would have been meticulously annotated by clinical experts (e.g., experienced gastroenterologists or endoscopists) to label polyps, potentially with subsequent pathological confirmation for cases where tissue was removed. This common practice ensures high-quality ground truth for training medical AI models.

Ask a Question

Ask a specific question about this device

Page 1 of 1