Multimodal Convolutional Deep Belief Networks for Stroke Classification with Fourier Transform

Mateus Roder, Nicolas Gomes, Arissa Yoshida, Fumie Costen, Joao Papa

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Several studies have investigated the vast potential of deep learning techniques in addressing a wide range of applications, from recommendation systems and service-based analysis to medical diagnosis. However, even with the remarkable
results achieved in some computer vision tasks, there is still a vast scope for exploration. Over the past decade, various studies focused on developing automated medical systems to support diagnosis. Nevertheless, detecting cerebrovascular accidents remains a challenging task. In this regard, one way to improve these approaches is to incorporate information fusion techniques in deep learning architectures. This paper proposes a novel approach to enhance stroke classification by combining multimodal data from Fourier transform with Convolutional Deep Belief Networks. As the main result, the proposed approach achieved state-of-the-art results with an accuracy of 99.94%, which demonstrates its effectiveness and potential for future applications.
Original languageEnglish
Title of host publication36th Conference on Graphics, Patterns and Images
Publication statusAccepted/In press - 4 Aug 2023


Dive into the research topics of 'Multimodal Convolutional Deep Belief Networks for Stroke Classification with Fourier Transform'. Together they form a unique fingerprint.

Cite this