A Neutrosophic Topological Algebraic Structure for Clinical Efficacy Evaluation in Traditional Chinese Medicine Using Artificial Intelligence

Abstract

The clinical evaluation of Traditional Chinese Medicine (TCM) presents unique challenges due to 
the inherent linguistic ambiguity of diagnostic criteria, the contextual variability of herbal 
prescriptions, and the pervasive uncertainty in patient response data. This study proposes a novel 
Neutrosophic Topological–Algebraic Framework that integrates two complementary components: 
(1) a Linguistic-Neutrosophic Clinical Topology (LNCT) for representing patient states and 
therapeutic outcomes as ordered neutrosophic triples (T, I, F) over a linguistic term set, structured 
by semi-topological axioms to preserve separability and regularity under uncertainty; and (2) a 
Soft Neutrosophic Quasigroup Interaction Algebra (SNQIA) to model context-dependent 
compositional operations on herbal combinations, ensuring algebraic closure and contextual 
adaptability. An evidential-neutrosophic neural network architecture is introduced, incorporating 
semi-T regularization and SNQIA-consistency constraints into its loss function to enforce both 
topological and algebraic integrity. 
The framework is validated on real-world TCM clinical datasets, including chronic low back pain 
and functional dyspepsia cases, where conventional probabilistic models struggle with linguistic 
vagueness and prescription variability. Experimental results demonstrate superior performance in 
Linguistic-Neutrosophic Efficacy Index (L-NEI) gain, improved calibration of indeterminacy, 
enhanced topological separability, and robustness to prescription-structure transformations 
representing differences across TCM schools. These findings suggest that the proposed approach 
provides a mathematically principled and computationally effective pathway toward interpretable 
and uncertainty-resilient clinical efficacy evaluation in TCM.

DOI: 10.5281/zenodo.16934490