<p>Within computer vision, the examination and understanding of medical images emerge as a crucial and captivating domain. The objective of our study is to build a system of intelligent recognition of biomedical images based on optimized fractional quaternion Gaussian–Hermite moments (FrQGHMs) by GWO algorithm and Deep Learning (DL). Our principal focus is on applications in the realm of medicine, especially in the analysis of medical images. For this purpose, we introduce a novel diagnosis intelligent system based on biomedical images through two steps: In the first step, we use a new image feature extraction technique based on a new set of optimized orthogonal moments by GWO algorithm. In the second step, we use the DNN deep learning method for image classification and prediction. The proposed method employs optimized fractional quaternion moments as a feature descriptor to extract features from biomedical images, these images have been consistently utilized and archived for both diagnostic and research purposes. Consequently, numerical experiments are carried out to demonstrate the effectiveness of the newly introduced FrQGHMs, comparing them to established methods, and recent fractional techniques. The results obtained from all the experiments carried out in this work show that our approach will open promising new horizons in the field of computer-aided medical diagnosis and recognition of cancerous diseases or other serious illnesses.</p>