tsta

فارسی

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Classification of Excitatory and Inhibitory Neurons in Animal Data Using Machine Learning and CNN Models

Authors :

Mahdi Mollaei¹ Amirhossein Mashghdoust² Ali Khadem³

1- دانشگاه صنعتی خواجه نصیرالدین طوسی تهران 2- دانشگاه صنعتی خواجه نصیرالدین طوسی تهران 3- دانشگاه صنعتی خواجه نصیرالدین طوسی تهران

Keywords :

Machine Learning،Deep Learning،Wavelet Transform،Neuron Classification،Hippocampal Mouse Data

Abstract :

Abstract— Automatically distinguishing excitatory from inhibitory (E/I) neurons in extracellular recordings is a fundamental challenge in neuroscience, as it enables accurate circuit dissection and more reliable spike sorting outcomes. In this study, we developed and evaluated models for cell-type classification based on extracellular action potentials (EAPs). Hippocampal recordings from the Buzsáki Lab, including 26 opto-tagged inhibitory neurons and matched excitatory neurons, were used to benchmark performance. Three classifiers were implemented: Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Convolutional Neural Networks (CNNs), trained on either raw spike waveforms or their two-dimensional wavelet transforms. A leave-one-out cross-validation scheme was applied to assess generalization. Among all tested models, CNNs achieved the highest accuracy (94.32%), outperforming both SVM (~92%) and LDA (~87%). Interestingly, CNNs performed slightly better on raw waveforms compared with wavelet inputs. While our results are slightly below those reported for CNNs on simulated data (~99%), they highlight the robustness of deep learning approaches when applied to real neuronal datasets. This work supports the growing evidence that deep networks can automatically extract informative features from extracellular recordings without hand-crafted metrics.