Many MRI/CT/X-Ray image compression techniques have been proposed for the past few years, but none of them highlighted image compression related to Hydrocephalus disease. Motivated by this fact, this article presents an enhanced encoding–decoding methodology for lossy compression and decompression, specifically applied to the Hydrocephalus secondary dataset, which includes MRI, CT, and X-ray images from various sources and modalities. The proposed approach is divided into several steps: preprocessing of MRI, CT, and X-ray images; designing the modified autoencoder-decoder architecture; segregating the training and testing sets; and generating results. The performance of this method is evaluated both quantitatively and qualitatively. Comparative tables highlight the method's superiority over existing benchmark techniques, achieving state-of-the-art results. The compression ratio of the proposed method is 9.42 which outperforms existing benchmark compression-decompression algorithms. The paper's major contributions are the unique MRI/CT/X-Ray images required for image compression technique, efficient use of deep learning concepts regarding the compression-decompression algorithm, result from tabulation for excellent parametric performance, and noble application on Hydrocephalus disease. This strategy is poised to aid in developing a web or cloud-based application, facilitating the compression of large volumes of MRI, CT, and X-ray images for medical practitioners and radiologists, thereby serving society both directly and indirectly.