Chronic exposure to arsenic, a prevalent issue in regions dependent on private wells, poses severe health risks, including neurotoxic effects in children. This study addresses the need for advanced diagnostic tools to assess arsenic exposure at the cellular level. We introduce a novel image dataset featuring phase contrast images of PC12 cells, which have been exposed to various concentrations of sodium arsenite in vitro, simulating the cellular impact of arsenic. To facilitate the analysis of this dataset, we developed an advanced convolutional neural network (CNN) model that incorporates the ResNet-50 architecture enhanced with a Convolutional Block Attention Module (CBAM). This model excels in identifying subtle morphological changes in cells, crucial for accurate arsenic concentration classification. We make the dataset and the CNN model publicly available to support further research and development in the field. This study provides a valuable tool for medical research communities working to mitigate the health impacts of arsenic exposure and which could be applied in the future to investigate any disease that affects neuron morphology.