.png){kind=icon}
.png){kind=icon}
Return
Whale optimization-based Lupus nephritis image segmentation: an adaptive approach with stable balancing
.png){kind=icon}
PRE
AI
Y
W
A
H
X
DOI：10.1007/s10586-026-06303-6.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Lupus nephritis (LN) is a severe renal complication, and the precise analysis of kidney images can significantly support clinical assessment. Therefore, accurate image segmentation of Lupus nephritis (LN) is essential for evaluating renal damage and guiding clinical interventions. However, the complex tissue structures in LN images, sourced from the electronic medical records of clinical patients at the Affiliated Hospital of Wenzhou Medical University in China, make precise segmentation highly challenging. While multi-threshold image segmentation (MIS) combined with meta-heuristic algorithms (MAs) is a promising approach for clinical diagnosis, traditional MAs frequently suffer from premature convergence and get trapped in local optima, resulting in suboptimal segmentation accuracy. To address this critical issue, this paper proposes an enhanced Whale Optimization Algorithm (ACNMWOA) specifically designed for robust MIS. The problem is solved by integrating three key mechanisms into the original WOA: a chaotic map during initialization to enhance early population diversity, an adaptable horizontal and vertical crossover strategy to boost global exploration and escape local optima, and the Nelder-Mead simplex (NMs) strategy to refine local exploitation for high-quality solutions. The effectiveness of the proposed method was rigorously evaluated. On the IEEE CEC 2017 benchmark, ACNMWOA demonstrated superior global optimization capabilities and effectively avoided local optima across various dimensions. In practical LN image segmentation applications, the proposed ACNMWOA-based MIS method significantly enhanced segmentation performance. Statistical results revealed that ACNMWOA outperformed other state-of-the-art methods in over 96% of the comparisons per threshold level at high thresholds and over 75% at low thresholds, proving its high accuracy and reliability for clinical LN image analysis.
Keywords:
Lupus Nephritis
Meta-heuristic algorithms
Whale Optimization Algorithm
Nelder-Mead simplex
Multi-threshold image segmentation
Journal
C
IF:
4.1
Papers:
4.8K
Citations:
7.5K
Organization
Cited Papers