Manuscript received February 22, 2024; revised May 20, 2024; accepted June 14, 2024; published January 17, 2025.

Abstract—Image processing in the field of modern health care has obtained a significant impact among the medical practitioners in aiding early diagnosis of diseases, treatment planning and disease monitoring. Neural Networks has been showing impressive success in many image analysis tasks. Convolutional Neural Networks (CNNs) framework has been exploited abundantly in medical image processing, in which the choice between two-Dimensional (2D) and three- Dimensional (3D) CNN architectures remains an open question in medical imaging applications. This research has investigated the difference in performance analysis between two-dimensional and three-dimensional CNNs for medical image classification tasks, focusing on accuracy, model complexity, and computational efficiency. Three-dimensional volumetric data of medical ultrasound images had been obtained, pre-processed using appropriate techniques and was given as input to CNN’s. Results indicated that 3D CNNs outperformed 2D CNNs in terms of classification accuracy, especially when the image data contained rich 3D spatial information. The 3D CNNs were able to capture spatial relationships and patterns across multiple slices, providing superior feature representations for medical image analysis. The architecture of 3D CNN was further enhanced by adding convolution layers so that it can read the patterns and voxel relationship with better accuracy than the traditional methods.

Keywords—2D-convoultional neural network, 3Dconvoultional neural network, 2D-image representation, 3Dimage representation, medical imaging, ultrasound image processing

Cite: Angelin Beulah S and Sivagami M, "Comparative Analysis of 2D and 3D Convolutional Neural Networks for Medical Ultrasound Image Classification," Journal of Image and Graphics, Vol. 13, No. 1, pp. 1-14, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.