EUVIMEDSkopri EUVIMED
Crossref

Modified Expectation Maximization Algorithm for Segmentation of Wireless Capsule Endoscopy Images in Absence of Ground Truth

Bhavin Dabhi, Mehul Parmar, Kuntesh Jani, Purvi Ramanuj, Subodh Srivastava

International Journal of Imaging Systems and Technology · 2026-07 · 36 (4)

DOI: 10.1002/ima.70420 ↗

At a glance

Journal article in International Journal of Imaging Systems and Technology, published online on 2026-07-15. Written by Bhavin Dabhi and 4 co-authors. 14 references.

AstrattDettalji biblijografiċiCiteRelated studies

Abstract (original language)

Capsule endoscopy is a noninvasive imaging technique that plays a crucial role in diagnosing abnormalities in the gastrointestinal (GI) tract. The large number of images generated during capsule endoscopy makes analysis and diagnosis challenging and time‐consuming. In this paper, we propose a novel Simulated Annealing—Expectation Maximization (SA‐EM) model for effective segmentation of GI tract features in capsule endoscopy images. The SA‐EM model improves upon the EM algorithm by addressing its limitation of getting stuck in local minima. The model explores the solution space by initially accepting worse solutions at higher temperatures and refining them over time to achieve optimal segmentation. Various segmentation techniques, such as modified U‐Net with scSE attention, Separable Contextual Feature Guided models, and SegNet, are reviewed and compared. The proposed SA‐EM model is evaluated on the Hyperkvasir dataset, showing superior performance with the highest mean Intersection over Union (mIoU) of 94.34%, a mean Dice score of 97%, and reduced uncertainty, demonstrated by the lowest mean Variation of Information (mVI) of 40.08% and mean Generalized Consistency Error (mGCE) of 5.58%. These results highlight the effectiveness of the SA‐EM model in improving diagnostic accuracy and minimizing segmentation uncertainty in capsule endoscopy images. Copyright 0000 John Wiley & Sons Ltd.

Cite

APA

Dabhi, B., Parmar, M., Jani, K., Ramanuj, P., & Srivastava, S. (2026). Modified Expectation Maximization Algorithm for Segmentation of Wireless Capsule Endoscopy Images in Absence of Ground Truth. International Journal of Imaging Systems and Technology, 36(4). https://doi.org/10.1002/ima.70420

Vancouver

Dabhi B, Parmar M, Jani K, Ramanuj P, Srivastava S. Modified Expectation Maximization Algorithm for Segmentation of Wireless Capsule Endoscopy Images in Absence of Ground Truth. International Journal of Imaging Systems and Technology. 2026;36(4). doi:10.1002/ima.70420

BibTeX
@article{Dabhi2026,
  author = {Dabhi, Bhavin and Parmar, Mehul and Jani, Kuntesh and Ramanuj, Purvi and Srivastava, Subodh},
  title = {Modified Expectation Maximization Algorithm for Segmentation of Wireless Capsule Endoscopy Images in Absence of Ground Truth},
  journal = {International Journal of Imaging Systems and Technology},
  year = {2026},
  volume = {36},
  number = {4},
  doi = {10.1002/ima.70420}
}
Start for free →