Two-stage machine learning models for bowel lesions characterisation using self-propelled capsule dynamics - research paper

Aug 4, 2026

Upload by Aurelia

Early bowel cancer diagnosis requires non-invasive biomechanical characterisation of bowel lesions. A two-stage machine learning framework is proposed that leverages the dynamics of a self-propelled capsule as it travels through the bowel and encounters lesions. Capsule measurements such as acceleration and displacement are extracted as signal features to reflect biomechanical differences, expressed via Young’s modulus. Supervised regression networks (MLP and SVR) predict Young’s modulus from dynamic features, followed by K-means clustering to group predicted moduli into distinct similarity-based clusters. Results show MLP outperforming SVR, and displacement-based models exceeding acceleration-based models when both signals are available.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Introduction

Bowel cancer background and screening methods

Capsule endoscopy and propulsion mechanisms

Study motivation and proposed approach

Nonlinear Dyn [https://doi.org/10.1007/s11071-023-08852-6](https://doi.org/10.1007/s11071-023-08852-6) ORIGINAL PAPER Two-stage machine learning models for bowel lesions characterisation using self-propelled capsule dynamics Kenneth Omokhagbo Afebu · Jiyuan Tian · Evangelos Papatheou · Yang Liu · Shyam Prasad Received: 9 March 2023 / Accepted: 18 August 2023 © The Author(s) 2023 Abstract To foster early bowel cancer diagnosis, anon-invasive biomechanical characterisation of bowel lesions is proposed. This method uses the dynamics of a self-propelled capsule and a two-stage machine learning procedure. As the capsule travels and encounters lesions in the bowel, its exhibited dynamics are envisaged to be of biomechanical signiﬁcance being a highly sensitive nonlinear dynamical system. For this study, measurable capsule dynamics including acceleration and displacement have been analysed for features that may be indicative of biomechanical differences, Young’s modulus in this case. The ﬁrst stage of the machine learning involves the development of supervised regression networks including multi-layer perceptron (MLP) and support vector regression (SVR), that are capable of predicting Young’s moduli from dynamic signals features. The second stage involves an K. O. Afebu · J. Tian · E. Papatheou · Y. Liu (B) Engineering Department, University of Exeter, Exeter EX4 4QF, UK e-mail: [y.liu2@exeter.ac.uk](y.liu2@exeter.ac.uk) K. O. Afebu[e-mail: k.afebu@exeter.ac.uk](e-mail: k.afebu@exeter.ac.uk) J. Tian e-mail: [e.papatheou@exeter.ac.uk](e.papatheou@exeter.ac.uk) E. Papatheou e-mail: [jt535@exeter.ac.uk](jt535@exeter.ac.uk) S. Prasad Royal Devon University Healthcare NHS Foundation Trust, Barrack Road, Exeter EX2 5DW, UK e-mail: [shyamprasad@nhs.net](shyamprasad@nhs.net) unsupervised categorisation of the predicted Young’s moduli into clusters of high intra-cluster similarity but low inter-cluster similarity using K-means clustering. Based on the performance metrics including coefﬁcient of determination and normalised mean absolute error, the MLP models showed better performanceson the test data compared to the SVR. For situations where both displacement and acceleration were measurable, the displacement-based models outperformed the acceleration-based models. These results thus make capsule displacement and MLP network the ﬁrst-line choices for the proposed bowel lesion characterisation and early bowel cancer diagnosis. Keywords Bowel cancer · Self-propelled capsule · Biomechanical properties · Signal analysis · Machine learning 1 Introduction Bowel cancer (BC), also referred to as colorectal cancer, affects the large bowel which consists ofboth colon and rectum. It is widely believed to have emerged from the adenoma-carcinoma sequence during which benign (i.e., adenoma) lesions mutate to become malignant (i.e., adenocarcinoma) and might also spread to other parts of the body, like the liver or lungs. BC ranks asthe second most deadliest cancer accounting for about a million deaths globally [22] with new cases expected to reach 3 .2 million in 2040 [67] . In the UK, about 1 3 268,000 persons are currently living with BC while about 43,000 new cases and 16,500 deaths are recorded every year. 94% of these new cases are diagnosed in people over the age of 50 years while the remaining 6% amounts to about 2,600 cases in people under the age of 50 [9, 12]. In England, the ﬁve-year survival rate ofBC stood at about58.7%for diagnosis madebetween 2014-2018and followedupto2019[49]. Between2013 and 2017, England cancer treatment data showed that patients have 98%, 93%, 89% and 44% survival rate within a year of treatment, if diagnosis is made at Stage I, II, III and IV, respectively [6] . The afore-stated thus makes early diagnosis very crucial to BC treatment and survival. Several methods are currently employed for BC screening, and these include rectal examination, faecal immunochemical test, comple

FAQs

What is the main goal of the proposed method?

To support early bowel cancer diagnosis by performing non-invasive biomechanical characterisation of bowel lesions using measurable capsule dynamics.

How does the two-stage machine learning pipeline work?

The first stage uses supervised regression (MLP or SVR) to predict Young’s modulus from capsule dynamic signal features. The second stage clusters predicted Young’s moduli using K-means based on within-cluster similarity and between-cluster separation.

Which model and capsule signal perform best?

MLP models show better performance than SVR on test data. When both displacement and acceleration are measurable, displacement-based models outperform acceleration-based models.

Aug 4, 2026

Upload by Aurelia