Artificial Intelligence and Polyp Differentiation

Artificial Intelligence and Polyp Differentiation

Thomas Rösch, Hamburg

GUT 2017 Oct 24. [Epub ahead of print]

Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model
Michael F Byrne1, Nicolas Chapados2,3, Florian Soudan2, Clemens Oertel2, Milagros Linares Pérez4, Raymond Kelly5, Nadeem Iqbal6, Florent Chandelier2, Douglas K Rex7


In general, academic but not community  endoscopists have demonstrated adequate endoscopic differentiation accuracy to make the ’resect and discard’ paradigm for diminutive colorectal polyps workable. Computer analysis of video could potentially eliminate the obstacle of interobserver variability in endoscopic polyp interpretation and enable widespread acceptance of ’resect and discard’.

Study design and methods

We developed an artificial intelligence (AI) model for real-time assessment of endoscopic video images of colorectal polyps. A deep convolutional neural network model was used. Only narrow band imaging video frames were used, split equally between relevant multiclasses. Unaltered videos from routine exams not specifically designed or adapted for AI classification were used to train and validate the model. The model was tested on a separate series of 125 videos of consecutively encountered diminutive polyps that were proven to be adenomas or hyperplastic polyps.


The AI model works with a confidence mechanism and did not generate sufficient confidence to predict the histology of 19 polyps in the test set, representing 15% of the polyps. For the remaining 106 diminutive polyps, the accuracy of the model was 94% (95% CI 86% to 97%), the sensitivity for identification of adenomas was 98% (95% CI 92% to 100%), specificity was 83% (95% CI 67% to 93%), negative predictive value 97% and positive predictive value 90%.


An AI model trained on endoscopic video can differentiate diminutive adenomas from hyperplastic polyps with high accuracy. Additional study of this programme in a live patient clinical trial setting to address resect and discard is planned.

Gastroenterology. 2017 Oct 14. [Epub ahead of print]

Accurate Classification of Diminutive Colorectal Polyps Using Computer-aided Analysis.
Peng-Jen Chen,1 Meng-Chiung Lin,2,3 Mei-Ju Lai,4 Jung-Chun Lin,1 Henry Horng-Shing Lu,Vincent S. Tseng6

Background & Aims

Narrow-band imaging (NBI) is an image-enhanced form of endoscopy used to observed microstructures and capillaries of the mucosal epithelium that allows for real-time prediction of histologic features of colorectal polyps. However, NBI expertise is required to differentiate hyperplastic from neoplastic polyps with high levels of accuracy. We developed and tested a system of computer-aided diagnosis with a deep neural network (DNN-CAD) to analyze narrow-band images of diminutive colorectal polyps.


We collected 1476 images of neoplastic polyps and 681 images of hyperplastic polyps, obtained from the picture archiving and communications system database in a tertiary hospital in Taiwan. Histologic findings from the polyps were also collected and used as the reference standard. The images and data were used to train the DNN. A test set of images (96 hyperplastic and 188 neoplastic polyps, smaller than 5 mm), obtained from patients who underwent colonoscopies from March 2017 through August 2017, was then used to test the diagnostic ability of the DNN-CAD vs endoscopists (2 expert and 4 novice), who were asked to classify the images of the test set as neoplastic or hyperplastic. Their classifications were compared with findings from histologic analysis. The primary outcome measures were diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic time. The accuracy, sensitivity, specificity, PPV, NPV, and diagnostic time were compared among DNN-CAD, the novice endoscopists, and the expert endoscopists. The study was designed to detect a difference of 10% in accuracy by a 2-sided McNemar test.


In the test set, the DNN-CAD identified neoplastic or hyperplastic polyps with 96.3% sensitivity, 78.1% specificity, a PPV of 89.6%, and a NPV of 91.5%. Fewer than half of the novice endoscopists classified polyps with a NPV of 90% (their NPVs ranged from 73.9% to 84.0%). DNN-CAD classified polyps as neoplastic or hyperplastic in 0.45±0.07 sec—shorter than the time required by experts (1.54±1.30 sec) and nonexperts (1.77±1.37 sec) (both P<.001). DNN-CAD classified polyps with perfect intra-observer agreement (kappa score of 1). There was a low level of intra-observer and inter-observer agreement in classification among endoscopists.


We developed a system called DNN-CAD to identify neoplastic or hyperplastic colorectal polyps less than 5 mm; it classified polyps with a PPV of 89.6%, and a NPV of 91.5%, and in a shorter time than endoscopists. This deep-learning model has potential for not only endoscopic image recognition but for other forms of medical image analysis, including sonography, computed tomography, and magnetic resonance images.


What you should know about these papers

Polyp differential diagnosis by endoscopy has become a hot topic in the frame of the DISCARD discussion, stating that histologic analysis of small/diminutive (< 5 mm) polyps is not necessary and that follow-up intervals should be determined by the endoscopic aspect1; some call this “endoscopic histology”.  Accordingly, guidelines have been made which accuracy levels and which levels of confidence should be reached in order to put this concept into clinical reality 2. Therafter, however, studies have appeared which show that this may work in highly motivated academic centers with a strong wish to publish excellent results, but much less so in clinical practice3-6; among those studies the largest one came from the UK and involved the inventors of DISCARD3.

This discrepancy may not speak against the principle itself but should prompt attempts to eliminate the subjective element of image interpretation and experience. Thus, the application of artificial intelligence (AI) appears logical in this respect. Nevertheless, it has to be made sure that any AI system is trained on the basis of validated images with histology being assessed by one (or better two) experienced histopathiologists (in consensus). In the first paper from Doug Rex and coworkers7 this was done to a limited extent only (videos with standard recording protocol, routine histopatholopgy), the Japanese study8 had a better histopathologic assessment (consensus reading), but a n artificially optimized image assessment (selected images of best quality); the latter may not represent daily work with moving images. Both papers had adequate methodology with training and validation sets.

In both papers, accuracy rates were around 80-95% with notable differences between sensitivity (98% US 96% Japan) and specificity (83% US, 78% Japan), mostly as good as with expert assessment in previous as well as in these papers2. This is probably quite good given the early development phase and the potential of deep learning to continuously improve its own results. The authors give account to these limitations and future prospects in the Discussions.

In general, there are even more exciting perspectives for AI in endoscopy, starting with polyp differentiation, assessment of other structures with regards to histology (Barrett, gastric abnormalities, IBD etc) which are surely more complex. With lesions characterization in mind, one should however go further to lesion detection, which is also a wide and potentially more difficult field. However, the automatic ADR (adenoma detection rate) optimized would be a fantastic tool to relieve the endoscopists from having always to be superconcentrated, and also better exploring the potential of wide angle scopes – could it be that the first generations such as the FUSE system (unfortunately not available any more at present) or an Olympus prototype could not explore their full potential due to human failures – and an automatic polyp detector would see everthing on multiple monitors (FUSE) or in a wide angle view (EWAVE) ? Let´s see, what comes next.



  1. Ignjatovic A, East JE, Suzuki N, et al. Optical diagnosis of small colorectal polyps at routine colonoscopy (Detect InSpect ChAracterise Resect and Discard; DISCARD trial): a prospective cohort study. Lancet Oncol 2009;10:1171-8.
  2. Abu Dayyeh BK, Thosani N, Konda V, et al. ASGE Technology Committee systematic review and meta-analysis assessing the ASGE PIVI thresholds for adopting real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointest Endosc 2015;81:502.e1-502.e16.
  3. Rees CJ, Rajasekhar PT, Wilson A, et al. Narrow band imaging optical diagnosis of small colorectal polyps in routine clinical practice: the Detect Inspect Characterise Resect and Discard 2 (DISCARD 2) study. Gut 2017;66:887-895.
  4. Vu HT, Sayuk GS, Hollander TG, et al. Resect and discard approach to colon polyps: real-world applicability among academic and community gastroenterologists. Dig Dis Sci 2015;60:502-8.
  5. Schachschal G, Mayr M, Treszl A, et al. Endoscopic versus histological characterisation of polyps during screening colonoscopy. Gut 2014;63:458-65.
  6. Ladabaum U, Fioritto A, Mitani A, et al. Real-time optical biopsy of colon polyps with narrow band imaging in community practice does not yet meet key thresholds for clinical decisions. Gastroenterology 2013;144:81-91.
  7. Byrne MF, Chapados N, Soudan F, et al. Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model. Gut 2017.
  8. Chen PJ, Lin MC, Lai MJ, et al. Accurate Classification of Diminutive Colorectal Polyps Using Computer-aided Analysis. Gastroenterology 2017.

Related Posts

Scroll to Top