Hyperspectral imaging can provide detailed information about the composition and chemical properties of objects across a wide range of the electromagnetic spectrum. These images capture intricate details and are used in medical imaging to identify cancerous regions in the brain. However, the large number of spectral bands in these images poses a challenge known as the “curse of dimensionality” during processing. To address this, feature selection and extraction methods can be employed. This study investigated the use of biogeography-based optimization for feature selection from hyperspectral images. Convolutional neural network techniques were then utilized to classify the selected bands, allowing for a quantitative assessment of the band selection methods. Extensive experiments were conducted using various CNN-based architectures to classify the selected bands, and the results reveal promising performance in hyperspectral brain image classification.