Digestive Endoscopy
Can we shorten the small-bowel capsule reading time with the “Quick-view” image detection system?

https://doi.org/10.1016/j.dld.2011.12.021Get rights and content

Abstract

Background

The mean small-bowel capsule reading time is about 60 min, and shortening this reading time is a major aim.

Aim

To evaluate the efficiency of the “Quick-view” detection algorithm.

Methods

Multicentre prospective comparative study. One hundred and six small bowel capsule films from 12 centres reviewed in Quick-view mode by 12 experienced readers. Reading time, image relevance, and comparison of Quick-view reading results to results of initial reading. Review of discordant result by 3 experts.

Results

The mean reading time in Quick-view mode was of 11.6 min (2–27). Concordant negative results were obtained in 41 cases (38.6%) and concordant positive results in 35 cases (33.0%). A discordant result was obtained in 30 (28.3%) cases: 21 false positive cases (initial reading 12 cases, Quick-view reading 9 cases), 14 false negative cases (initial reading 7, Quick-view 7). Four out of 7 lesions missed at Quick-view reading were not present on the Quick-view film (theoretical sensitivity 93.5%).

Conclusion

The Quick-view informatic algorithm detected nearly 94% of significant lesions, and Quick-view reading was as efficient as the initial reading and much shorter. These results are to be confirmed by further studies, but suggest an excellent sensitivity for the Quick-view algorithm.

Introduction

Capsule endoscopy is a first choice small bowel examination for most gastroenterologists worldwide in patients presenting with obscure (overt or occult) gastrointestinal bleeding (OGIB), because of simple use, low risk, and high sensitivity as compared to all other small-bowel imaging modalities [1], [2]. Capsule endoscopy is a very sensitive examination for the detection of responsible small bowel lesions, as assessed by the long term follow-up of patients after negative capsule endoscopy: only 10% of patients present with ongoing bleeding related to a missed small-bowel lesion [3]. The main limitation of capsule endoscopy is the substantial time needed for capsule reading by highly qualified medical personnel. In most series [4], [5], [6], a mean reading time of 50–60 min is reported. In addition, the focus and attention to detail that is necessary to efficiently read a film is difficult to maintain during the whole reading [4], [7]. For this reason, attempts have been made to transfer capsule reading to nurses or other similarly qualified personnel, with relatively satisfying results [8], [9]. Complex informatic approaches to solving the reading time dilemma have also been proposed, with variable results. This approach needs clinical evaluation and validation [10], [11].

Clinically validated diagnostic algorithms targeting and selecting for viewing of significant images from what is on average an approximately about 64 800 image videos (for the Given Imaging° capsule) are thus of major importance for gastroenterology practice. The first informatic algorithm developed for assisting in lesion detection during capsule reading was the SBI (small bowel blood indicator, Given Imaging°). The SBI system was designed to detect all red lesions potentially responsible for anaemia or bleeding. However, this detection system proved to be of poor sensitivity and specificity [12], [13]. The Quick-view (QV) algorithm is a new image-detection system, also developed by Given Imaging. The potential of this informatic algorithm is to reduce drastically the reading time, and perhaps to facilitate the detection of significant small bowel lesions missed during the normal reading of capsule films. The Quick view algorithm is, however, inadequately validated: at the present time, only one peer-reviewed study evaluated this algorithm, with rather unsatisfying results [14]. We thus performed a prospective, muticentre evaluation of this algorithm in unselected films from ten French gastroenterology centres, with the aim of evaluating the sensitivity and specificity of a QV reading compared to the previous normal reading of the capsule film.

Section snippets

Films

The study includes 106 small bowel capsule films from 106 unselected patients corresponding to the last 5–10 cases from each of the 10 experienced French centres belonging to the French Society of Digestive Endoscopy (SFED).

The capsule endoscopy (PillCam SB 2°, Given Imaging, Yokneam, Israel) procedure had been performed in each centre following different local preparation protocols following the local usual procedure. Each capsule film had been read initially by one experienced local reader.

Results

The 106 CE films were read in QV mode without technical issues, with a mean reading time of 11.6 (range 2–27) min.

Discussion

We show in the present investigation that the QV algorithm may represent an important advance for capsule endoscopy film reading by providing a rapid (a mean of 11.6 min) and efficient (89% sensitivity, exactly comparable to a usual complete PillCam SB2° film reading) [3], [7]. Evaluating new detection or reading algorithms is a highly important process as the quality of patients’ management depends on the efficiency of the algorithm. Decreasing reading time is one of the future major

Conflict of interest

JCS, BF and SSH work as occasional consultants for Given Imaging.

References (20)

There are more references available in the full text version of this article.

Cited by (52)

  • The evolving role of small-bowel capsule endoscopy

    2023, Best Practice and Research: Clinical Gastroenterology
  • Colon capsule endoscopy and artificial intelligence: a perfect match for panendoscopy

    2023, Artificial Intelligence in Capsule Endoscopy: A Gamechanger for a Groundbreaking Technique
  • Striving for quality improvement: can artificial intelligence help?

    2021, Best Practice and Research: Clinical Gastroenterology
  • Training, Reading, and Reporting for Small Bowel Video Capsule Endoscopy

    2021, Gastrointestinal Endoscopy Clinics of North America
  • Deep learning for wireless capsule endoscopy: a systematic review and meta-analysis

    2020, Gastrointestinal Endoscopy
    Citation Excerpt :

    It is worth noting that work has also been conducted for WCE lesion detection using classic machine learning.55-58 Several commercial machine learning systems have been proposed, including the Suspected Blood Indicator algorithm (Medtronic),59 the Quick-View algorithm (Medtronic),60 and the ExpressView algorithm (IntroMedic).61 A number of studies included in this review have compared CNN performance with prior machine learning methods and have shown that the CNN outperformed the latter.34,38,41,62-65

View all citing articles on Scopus
View full text