A Systematic Analysis of the Impact of BloodMNIST Data Characteristics on CNN Classification Performance
DOI:
https://doi.org/10.54097/d22zbt09Keywords:
BloodMNIST, convolutional neural network, classification performance, misclassification analysis, visual similarity, class imbalance, noise robustnessAbstract
BloodMNIST, as a standardized medical image classification dataset, has been widely used to validate the classification performance of convolutional neural networks (CNNs). This paper takes BloodMNIST as the research subject, focusing on the key data characteristics that affect CNN classification performance and typical misclassified samples. Through comparative experiments on the classification results of LeNet and ResNet, we systematically investigate the mechanisms by which different data properties influence model behavior. Specifically, this study conducts experiments around three core influencing factors: (1) class visual similarity through sample visualization and confusion matrix analysis, we reveal the impact of similar classes on misclassification; (2) class sample size (class imbalance) – by comparing standard cross‑entropy with weighted cross‑entropy, we analyze the effect of sample distribution on classification performance; (3) sample quality/noise – by introducing mild perturbations, we evaluate the robustness of different models. Experimental results show that visual similarity and class imbalance are important causes of misclassification, and ResNet is more robust than LeNet in distinguishing similar classes and resisting noise. This paper provides a systematic experimental analysis framework for understanding the relationship between medical image data characteristics and CNN classification behavior.
Downloads
References
[1] Litjens, G., Kooi, T., Bejnordi, B. E., et al. (2017). A survey on deep learning in medical image analysis. Medical Image Analysis, 42, 60 88.
[2] Szegedy, C., Vanhoucke, V., Ioffe, S., et al. (2016). Rethinking the inception architecture for computer vision. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 2818 2826). IEEE.
[3] Hendrycks, D., & Gimpel, K. (2017). A baseline for detecting misclassified and out of distribution examples in neural networks. In Proceedings of the International Conference on Learning Representations (ICLR). ICLR.
[4] Yang, J., Shi, R., Wei, D., et al. (2023). MedMNIST v2: A large scale lightweight benchmark for 2D and 3D biomedical image classification. Scientific Data, 10(1), 41.
[5] LeCun, Y., Bottou, L., Bengio, Y., et al. (1998). Gradient based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278 2324.
[6] He, K., Zhang, X., Ren, S., et al. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 770 778). IEEE.
[7] He, H., & Garcia, E. A. (2009). Learning from imbalanced data. IEEE Transactions on Knowledge and Data Engineering, 21(9), 1263 1284.
[8] Cui, Y., Jia, M., Lin, T. Y., et al. (2019). Class balanced loss based on effective number of samples. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 9268 9277). IEEE.
[9] Goodfellow, I. J., Shlens, J., & Szegedy, C. (2015). Explaining and harnessing adversarial examples. In Proceedings of the International Conference on Learning Representations (ICLR). ICLR.
[10] Selvaraju, R. R., Cogswell, M., Das, A., et al. (2017). Grad CAM: Visual explanations from deep networks via gradient based localization. In Proceedings of the IEEE International Conference on Computer Vision (ICCV) (pp. 618 626). IEEE.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Academic Journal of Applied Sciences
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
