Registration is open - make plans to attend
Get pricing and details
>
Conference 12033 > Paper 12033-84
Paper 12033-84

Colorectal polyp classification using confidence-calibrated convolutional neural networks

In person: 23 February 2022 • 5:30 PM - 7:00 PM PST

Abstract

Computer-Aided Diagnosis (CADx) systems for in-vivo characterization of Colorectal Polyps (CRPs) which are precursor lesions of Colorectal Cancer (CRC), can assist clinicians with diagnosis and better informed decisionmaking during colonoscopy procedures. Current deep learning-based state-of-the-art solutions achieve a high classification performance, but lack measures to increase the reliability of such systems. In this paper, the reliability of a Convolutional Neural Network (CNN) for characterization of CRPs is specifically addressed by confidence calibration. Well-calibrated models produce classification-confidence scores that reflect the actual correctness likelihood of the model, thereby supporting reliable predictions by trustworthy and informative confidence scores. Two recently proposed trainable calibration methods are explored for CRP classification to calibrate the confidence of the proposed CNN. We show that the confidence-calibration error can be decreased by 33.86% (−0.01648 ± 0.01085), 48.33% (−0.04415 ± 0.01731), 50.57% (−0.11423 ± 0.00680), 61.68% (−0.01553 ± 0.00204) and 48.27% (−0.22074 ± 0.08652) for the Expected Calibration Error (ECE), Average Calibration Error (ACE), Maximum Calibration Error (MCE), Over-Confidence Error (OE) and Cumulative Calibration Error (CUMU), respectively. Moreover, the absolute difference between the average entropy and the expected entropy was considerably reduced by 32.00% (−0.04374 ± 0.01238) on average. Furthermore, even a slightly improved classification performance is observed, compared to the uncalibrated equivalent. The obtained results show that the proposed model for CRP classification with confidence calibration produces better calibrated predictions without sacrificing classification performance. This work shows promising points of engagement towards obtaining reliable and well-calibrated CADx systems for in-vivo polyp characterization, to assist clinicians during colonoscopy procedures.

Presenter

Technische Univ. Eindhoven (Netherlands)
Koen graduated with distinction cum laude from his master in Electrical Engineering at Eindhoven University of Technology last year and the work he will be presenting is about the work he did for his Masters thesis. Koen is fascinated by the application of machine learning and computer vision to clinical problems and in particular for oncology applications. This also motivated Koen to start a PhD project on the early detection of another type of gastrointestinal cancer, namely Esophageal cancer.
Presenter/Author
Technische Univ. Eindhoven (Netherlands)
Author
Technische Univ. Eindhoven (Netherlands)
Author
Technische Univ. Eindhoven (Netherlands)
Author
Maastricht Univ. Medical Ctr. (Netherlands), GROW, School for Oncology and Developmental Biology (Netherlands)
Author
Catharina Hospital (Netherlands)
Author
Maastricht Univ. Medical Ctr. (Netherlands)
Author
Erik J. Schoon
Catharina Hospital (Netherlands)
Author
Fons van der Sommen
Technische Univ. Eindhoven (Netherlands), Eindhoven Artificial Intelligence Systems Institute (Netherlands)
Author
Technische Univ. Eindhoven (Netherlands)