Abstract
Spiking Neural Networks (SNNs) event-driven nature enables efficient encodingof spatial and temporal features, making them suitable for dynamictime-dependent data processing. Despite their biological relevance, SNNs haveseen limited application in medical image recognition due to difficulties inmatching the performance of conventional deep learning models. To address this,we propose a novel breast cancer classification approach that combines SNNswith Lempel-Ziv Complexity (LZC) a computationally efficient measure ofsequence complexity. LZC enhances the interpretability and accuracy ofspike-based models by capturing structural patterns in neural activity. Ourstudy explores both biophysical Leaky Integrate-and-Fire (LIF) andprobabilistic Levy-Baxter (LB) neuron models under supervised, unsupervised,and hybrid learning regimes. Experiments were conducted on the Breast CancerWisconsin dataset using numerical features derived from medical imaging.LB-based models consistently exceeded 90.00% accuracy, while LIF-based modelsreached over 85.00%. The highest accuracy of 98.25% was achieved using anANN-to-SNN conversion method applied to both neuron models comparable totraditional deep learning with back-propagation, but at up to 100 times lowercomputational cost. This hybrid approach merges deep learning performance withthe efficiency and plausibility of SNNs, yielding top results at lowercomputational cost. We hypothesize that the synergy between temporal-coding,spike-sparsity, and LZC-driven complexity analysis enables more-efficientfeature extraction. Our findings demonstrate that SNNs combined with LZC offerpromising, biologically plausible alternative to conventional neural networksin medical diagnostics, particularly for resource-constrained or real-timesystems.