Chaotic Learning Rate Scheduling for Improved CNN-Based Breast Cancer Ultrasound Classification

Year 2025, Volume: 7 Issue: 3, 297 - 306, 30.11.2025

Ishak Pacal

https://doi.org/10.51537/chaos.1807694

https://izlik.org/JA29ZJ29NF

Abstract

Early and accurate identification of breast cancer from ultrasound is a key clinical objective, yet performance remains sensitive to how deep models are trained. Convolutional neural networks (CNNs) achieve strong results, but conventional learning rate schedulers follow rigid, deterministic trajectories that can limit exploration of complex loss landscapes. In this study, we introduce a Chaotic Learning Rate Scheduler (CLRS) that modulates the learning rate via the bounded, nonperiodic dynamics of the logistic map to encourage broader yet stable exploration during training. We assess CLRS in a controlled comparison on the BUSI dataset using four pretrained backbones. All components of the pipeline other than the scheduling policy are fixed, including data splits, preprocessing, optimizer settings, and model selection. Under an identical 300-epoch budget, CLRS is evaluated against a cosine scheduler, and test performance is reported using accuracy, precision, recall, and macro F1. LayerCAM is used to examine whether numerical gains correspond to clinically meaningful spatial attention. Across all backbones, CLRS consistently outperforms the cosine baseline. The best configuration, based on EfficientNetV2 Small, attains 0.9391 accuracy and 0.9255 macro F1. Gains are most pronounced in recall and macro F1, indicating improved sensitivity without additional inference cost. Parameter counts and GMACs remain unchanged, showing that benefits arise from altered training dynamics rather than model capacity. LayerCAM visualizations reveal more lesion-centric attention in correctly classified cases and support CLRS as an effective, deployment-neutral strategy for breast ultrasound classification.

Keywords

Breast cancer , Deep learning , Convolutional neural networks , Chaotic learning rate scheduler , Logistic map , Ultrasound

References

• Abufadel, A., G. Slabaugh, G. Unal, L. Zhang, and B. Odry, 2006 Interacting active rectangles for estimation of intervertebral disk orientation. In 18th International Conference on Pattern Recognition (ICPR’06), volume 1, pp. 1013–1016, IEEE.
• Al-Dhabyani, W., M. Gomaa, H. Khaled, and A. Fahmy, 2020a Dataset of breast ultrasound images. Data in brief 28: 104863.
• Al-Dhabyani, W., M. Gomaa, H. Khaled, and A. Fahmy, 2020b Dataset of breast ultrasound images. Data in Brief 28: 104863.
• Alswilem, L. and E. Asadov, 2025 Densenet-resnet-hybrid: A novel hybrid deep learning architecture for accurate apple leaf disease detection. Computational Systems and Artificial Intelligence 1: 1–7.
• Alswilem, L. and N. Pacal, 2025 Artificial intelligence in mammography: A study of diagnostic accuracy and efficiency. Computational Systems and Artificial Intelligence 1: 26–31.
• Aslan, E., S. D. Alpsalaz, F. Alpsalaz, H. Uzel, et al., 2025a Alzheimer’s classification with a maxvit-based deep learning model using magnetic resonance imaging. Journal of Applied Science and Technology Trends 6.
• Aslan, E. and Y. Özüpak, 2024 Advanced skin cancer detection using convolutional neural networks and transfer learning. Middle East Journal of Science 10: 167–178.
• Aslan, E. and Y. Özüpak, 2025 Comparison of machine learning algorithms for automatic prediction of alzheimer disease. Journal of the Chinese Medical Association 88: 98–107.
• Aslan, E., Y. Ozupak, F. Alpsalaz, and Z. M. Elbarbary, 2025b A hybrid machine learning approach for predicting power transformer failures using internet of things based monitoring and explainable artificial intelligence. IEEE Access .
• Attallah, O. and I. Pacal, 2026 Impact of magnification on deep learning approaches through comprehensive comparative study of histopathological breast cancer classification. Biomedical Signal Processing and Control 113: 108973.
• Burukanli, M. and N. Yumu¸sak, 2024a Stackgridcov: a robust stacking ensemble learning-based model integrated with gridsearchcv hyperparameter tuning technique for mutation prediction of covid-19 virus. Neural Computing and Applications 36: 22379–22401.
• Burukanli, M. and N. Yumu¸sak, 2024b Tfradmcov: a robust transformer encoder based model with adam optimizer algorithm for covid-19 mutation prediction. Connection Science 36: 2365334.
• Çakmak, Y., 2025 Machine learning approaches for enhanced diagnosis of hematological disorders. Computational Systems and Artificial Intelligence 1: 8–14.
• Cakmak, Y. and I. Pacal, 2025 Ai-driven classification of anemia and blood disorders using machine learning models. Computers and Electronics in Medicine 2: 43–52.
• Çakmak, Y. and N. Pacal, 2025 Deep learning for automated breast cancer detection in ultrasound: A comparative study of four cnn architectures. Artificial Intelligence in Applied Sciences 1: 13–19.
• Çakmak, Y. and J. Zeynalov, 2025 A comparative analysis of convolutional neural network architectures for breast cancer classification from mammograms. Artificial Intelligence in Applied Sciences 1: 28–34.
• Carrilero-Mardones, M., M. Parras-Jurado, A. Nogales, J. PérezMartín, and F. J. Díez, 2024 Deep learning for describing breast ultrasound images with bi-rads terms. Journal of Imaging Informatics in Medicine 37: 2940–2954.
• Chen, S., S. Feng, W. Fu, and Y. Zhang, 2021 Logistic map: Stability and entrance to chaos. In Journal of Physics: Conference Series, volume 2014, p. 012009, IOP Publishing.
• Chollet, F., 2017 Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1800–1807.
• He, K., X. Zhang, S. Ren, and J. Sun, 2016 Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778.
• Hosbas, M. Z., B. Emin, and F. Kaçar, 2025 True random number generator design with a fractional order sprott b chaotic system. ADBA Computer Science 2: 50–55.
• Hu, D., Z. Li, B. Zheng, X. Lin, Y. Pan, et al., 2022 Cancer-associated fibroblasts in breast cancer: Challenges and opportunities. Cancer Communications 42: 401–434.
• Huang, G., Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, 2017 Densely connected convolutional networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2261–2269.
• Ince, S., I. Kunduracioglu, A. Algarni, B. Bayram, and I. Pacal, 2025 Deep learning for cerebral vascular occlusion segmentation: a novel convnextv2 and grn-integrated u-net framework for diffusion-weighted imaging. Neuroscience 574: 42–53.
• ˙Ince, S., I. Kunduracioglu, B. Bayram, and I. Pacal, 2025 U-netbased models for precise brain stroke segmentation. Chaos Theory and Applications 7: 50–60.
• Jabeen, K., M. A. Khan, A. Hamza, H. M. Albarakati, S. Alsenan, et al., 2025 An efficientnet integrated resnet deep network and explainable ai for breast lesion classification from ultrasound images. CAAI Transactions on Intelligence Technology 10: 842–857.
• Johnson, O. V., C. Xinying, K. W. Khaw, and M. H. Lee, 2023 pscalr: periodic-shift cosine annealing learning rate for deep neural networks. IEEE Access 11: 139171–139186.
• Kaya, Y. and Z. G. Aydin, 2025 A chaos-based encryption scheme for secure medical x-ray images. Computers and Electronics in Medicine 2: 53–59.
• Kıran, H. E., 2025 Deep learning-based detection of abdominal diseases using yolov9 models and advanced preprocessing techniques. Computers and Electronics in Medicine 2: 20–25.
• Kumar, K., R. Pandey, S. S. Bhattacharjee, and N. V. George, 2021 Exponential hyperbolic cosine robust adaptive filters for audio signal processing. IEEE Signal Processing Letters 28: 1410–1414.
• Li, L., Y. Niu, F. Tian, and B. Huang, 2025a An efficient deep learning strategy for accurate and automated detection of breast tumors in ultrasound image datasets. Frontiers in Oncology 14: 1461542.
• Li, M., Y. Fang, J. Shao, Y. Jiang, G. Xu, et al., 2025b Vision transformer-based multimodal fusion network for classification of tumor malignancy on breast ultrasound: A retrospective multicenter study. International Journal of Medical Informatics 196: 105793.
• Liu, B., S. Liu, Z. Cao, J. Zhang, X. Pu, et al., 2025 Accurate classification of benign and malignant breast tumors in ultrasound imaging with an enhanced deep learning model. Frontiers in Bioengineering and Biotechnology 13: 1526260.
• Medghalchi, Y., N. Zakariaei, A. Rahmim, and I. Hacihaliloglu, 2025 Synthetic vs. classic data augmentation: Impacts on breast ultrasound image classification. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control .
• Ngongiah, I. K., R. T. Fotsa, J. M. Ngono, N. N. Kibanya, E. Godwe, et al., 2025 Chaos annihilation-based on genetic algorithms in a non-smooth air-gap permanent magnet synchronous motor embedded in the microcontroller. ADBA Computer Science 2:30–35.
• Orrantia-Borunda, E., L. E. A. Aguilar, and C. A. R. Valdespino, 2022 Nanomaterials for breast cancer. Exon Publications pp. 149–162.
• Pacal, I., A. Algarni, B. Bayram, and S. Ince, 2025 Fa-unet: A fasternet and attention-gated hybrid network for precise ischemic stroke segmentation. Journal of Integrative Neuroscience 24:40100.
• Pacal, I. and O. Attallah, 2025 Inceptionnext-transformer: A novel multi-scale deep feature learning architecture for multimodal breast cancer diagnosis. Biomedical Signal Processing and Control 110: 108116.
• Pacal, I. and Y. Cakmak, 2025 A comparative analysis of u-netbased architectures for robust segmentation of bladder cancer lesions in magnetic resonance imaging. Eurasian Journal of Medicine and Oncology p. 025260276.
• Pareek, N. K., V. Patidar, and K. K. Sud, 2006 Image encryption using chaotic logistic map. Image and vision computing 24: 926– 934.
• Qasrawi, R., O. Daraghmeh, S. Thwib, I. Qdaih, G. Issa, et al., 2025 Advancing breast cancer detection in ultrasound images using a novel hybrid ensemble deep learning model. Intelligence-Based Medicine 11: 100222.
• Siegel, R. L., T. B. Kratzer, A. N. Giaquinto, H. Sung, and A. Jemal, 2025 Cancer statistics, 2025. Ca 75: 10.
• Taheri, F. and K. Rahbar, 2025 Improving breast cancer classification in fine-grain ultrasound images through feature discrimination and a transfer learning approach. Biomedical Signal Processing and Control 106: 107690.
• Tan, M. and Q. Le, 2021 Efficientnetv2: Smaller models and faster training. In International conference on machine learning, pp. 10096– 10106, PMLR.
• Verma, K. and A. Maiti, 2025 Sine and cosine based learning rate for gradient descent method. Applied Intelligence 55: 352.
• Wang, J. and S.-G. Wu, 2023 Breast cancer: an overview of current therapeutic strategies, challenge, and perspectives. Breast Cancer: Targets and Therapy pp. 721–730.
• Wilkinson, L. and T. Gathani, 2022 Understanding breast cancer as a global health concern. The British journal of radiology 95: 20211033.
• Xu, Z., S. Zhong, Y. Gao, J. Huo, W. Xu, et al., 2025 Optimizing breast lesions diagnosis and decision-making with a deep learning fusion model integrating ultrasound and mammography: a dual-center retrospective study. Breast Cancer Research 27: 80.
• Yılmaz, M. T., E. Algul, and I. Pacal, 2025 A comparative study of advanced deep learning architectures for breast cancer classification on ultrasound and histological images. Results in Engineering p. 107600.
• Zagami, P. and L. A. Carey, 2022 Triple negative breast cancer:Pitfalls and progress. NPJ breast cancer 8: 95.
• Zhu, C., X. Chai, Z. Wang, Y. Xiao, R. Zhang, et al., 2024 Dbl-net: A dual-branch learning network with information from spatial and frequency domains for tumor segmentation and classification in breast ultrasound image. Biomedical Signal Processing and Control 93: 106221.