Layer-Wise Persistent Entropy of CT Scan Point Clouds for Lung Tumor Classification
AppliedMath · 2026-06-11 · 6 (6) · 95
DOI: 10.3390/appliedmath6060095 ↗At a glance
Journal article in AppliedMath, published online on 2026-06-11. Written by C. Jeeva Jose and 8 co-authors. 29 references. The full text is freely available (open access, CC BY 4.0). Funded by: APJ Abdul Kalam Technological University, Centre for Engineering Research and Development, Council of Scientific and Industrial Research (CSIR), India.
Abstract (original language)
In this study, a layer-wise point cloud representation of CT scan images is proposed, from which persistence diagrams are constructed and persistent entropy is computed as a compact topological feature for three-class lung tumor classification. Two parallel approaches are investigated: the direct computation of persistence diagrams from CT images, and computation from subsampled point clouds derived from image intensity layers. The proposed method is evaluated on the publicly available IQ-OTH/NCCD lung cancer dataset, comprising 1097 CT scan images from 110 individuals, annotated by expert oncologists and radiologists. Classification is performed using K-Nearest Neighbors (KNN), Random Forest, Support Vector Machine, and eXtreme Gradient Boosting, and compared against Convolutional Neural Network (CNN) and traditional image feature-based methods. The persistent entropy approach applied to layer-wise subsampled point clouds achieves 97.67% accuracy, a Precision–Recall AUC of 96.63%, and a ROC-AUC of 99.46% using KNN, outperforming direct image-based analysis (95.91%) and achieving comparable accuracy to the CNN method (97.21%) with a computational speedup of approximately 478×. These results demonstrate that persistent homology applied to subsampled point clouds provides an accurate, mathematically interpretable, and computationally efficient alternative to deep learning for lung tumor classification.
Funding
- APJ Abdul Kalam Technological University
- Centre for Engineering Research and Development
- Council of Scientific and Industrial Research (CSIR), India
Cite
APA
Jose, C. J., Baiju, A. P., Roy, R., Harikrishnan, A., Sanju, R., Kumar, P. B. V., Sherly, K. K., Jacob, R., & Sreekumar, G. (2026). Layer-Wise Persistent Entropy of CT Scan Point Clouds for Lung Tumor Classification. AppliedMath, 6(6), 95. https://doi.org/10.3390/appliedmath6060095
Vancouver
Jose CJ, Baiju AP, Roy R, Harikrishnan A, Sanju R, Kumar PBV, et al. Layer-Wise Persistent Entropy of CT Scan Point Clouds for Lung Tumor Classification. AppliedMath. 2026;6(6):95. doi:10.3390/appliedmath6060095
BibTeX
@article{Jose2026,
author = {Jose, C. Jeeva and Baiju, Aneesh P. and Roy, Riya and Harikrishnan, A. and Sanju, Rahul and Kumar, P. B. Vinod and Sherly, K. K. and Jacob, Rinku and Sreekumar, G.},
title = {Layer-Wise Persistent Entropy of CT Scan Point Clouds for Lung Tumor Classification},
journal = {AppliedMath},
year = {2026},
volume = {6},
number = {6},
pages = {95},
doi = {10.3390/appliedmath6060095}
}
EUVIMED