Abstract
In vitro modeling of neuronal disorders using transfected primary neurons is one of the fundamental approaches for studying disease mechanisms and therapeutic screening. Assessing neuronal viability is an everyday yet critical task in such experiments and requires accurate identification and classification of live and dead transfected neurons from dual-channel fluorescence images; however, this step is typically performed manually, resulting in inconsistent, labor-intensive, and poorly scalable analysis due to limitations of existing image analysis tools. Here, we present BlueNuclei, a user-friendly software with two modules: Hyades, which identifies nuclei of transfected neurons using dual-channel fluorescence image processing techniques, and Pleiades, an SVM-based classifier that distinguishes live from dead neurons using human-vision-inspired, biologically interpretable subnuclear features. Benchmarking on real images showed that BlueNuclei achieves near-human accuracy with substantially faster processing and minimal computational resources compared to deep learning alternatives when applied to the classification step. BlueNuclei provides a simple local user interface for data input and interactive visualizations that display classification results, including feature metrics and a confidence score for each nucleus. BlueNuclei offers the first scalable, fully automated, solution to viability assessment of transfected neurons, facilitating in vitro mechanistic studies of genetic neuronal disorders and therapeutic screening.