Neural diseases, such as stroke, affect the motor abilities of millions of people worldwide. Over the past few decades, numerous innovative neurorehabilitation methods have been developed to tackle this challenge. Brain-Computer Interfaces have emerged as powerful tools for neurological therapies, particularly with the advent of deep learning techniques utilizing Convolutional Neural Networks. This study proposes and evaluates an electroencephalographic signal processing method based on continuous Progressive Neural Network algorithm for decoding motor imagery during pedaling. Our approach demonstrates classification accuracy exceeding 80%, highlighting its potential for enhancing neurorehabilitation therapies, especially by mitigating the effects of catastrophic forgetting.