Unsupervised Clustering Methods for Lung Perfusion Data Segmentation in Electrical Impedance Tomography

EasyChair Preprint 15778

Abstract

In this work, we evaluated unsupervised clustering methods in segmenting the electrical impedance tomography image during the assessment of pulmonary perfusion by injection of hypertonic saline solution. In clustering the image pixels, we assume the existence of purely lung pixels (solely due to lung perfusion without effects from other organs) and hybrid pixels (which contain heart and lung effects together). We used data from 5 pigs to generate truth masks and do the proper clustering. Among the methods tested, the k-means with the cosine metric proved the best, as it obtained the lowest error of hybrid pixels wrongly classified as pulmonary (3%). This error is undesired, as it implies estimating pulmonary perfusion that actually comes from the heart signal. We prioritized minimizing such error as it would overestimate regional pulmonary perfusion.

Keyphrases: Clustering, Electrical Impedance Tomography, K-means, hierarchical clustering, lung perfusion

@booklet{EasyChair:15778,
  author    = {Arthur Ribeiro and Yu Xia and Monica Matsumoto and Marcus Victor},
  title     = {Unsupervised Clustering Methods for Lung Perfusion Data Segmentation in Electrical Impedance Tomography},
  howpublished = {EasyChair Preprint 15778},
  year      = {EasyChair, 2025}}