Abstract
Accurate differentiation between primary central nervous system lymphoma (PCNSL) and glioblastoma is essential for appropriate treatment selection, as these conditions have distinct therapeutic approaches and prognoses. Traditional diagnostic methods often rely on subjective visual assessment, which can lead to inconsistent results. In this work, we propose a sparse representation-based system for automated differentiation of PCNSL from glioblastoma.
Our approach leverages sparse representation theory to capture the underlying structure of brain tumor imaging patterns. The framework consists of three main components: (1) a sparse coding module that learns discriminative representations, (2) a dictionary learning mechanism that captures tumor-specific features, and (3) a classification module that provides automated differentiation.
We evaluate our method on a comprehensive dataset of PCNSL and glioblastoma cases with confirmed diagnoses. Experimental results demonstrate that our sparse representation approach significantly improves differentiation accuracy compared to traditional methods. The system shows robust performance across different imaging protocols and provides interpretable results.
The proposed framework represents a significant advancement in CNS lymphoma diagnosis, providing more accurate differentiation that could improve treatment planning and patient outcomes.
BibTeX
@article{chen2018primary,
title={Primary central nervous system lymphoma and glioblastoma differentiation based on conventional magnetic resonance imaging by high-throughput SIFT features},
author={Chen, Yinsheng and Li, Zeju and Wu, Guoqing and Yu, Jinhua and Wang, Yuanyuan and Lv, Xiaofei and Ju, Xue and Chen, Zhongping},
journal={International Journal of Neuroscience},
year={2018},
publisher={Taylor \& Francis},
doi={10.1080/00207454.2017.1408613}
}