A Comprehensive Review of Datasets and Preprocessing Techniques in Deep Learning-Based Breast Cancer Classification
Sharmin Akhter, Afjal Hossan Sarower, Farha Kamal, Tasmina Imam, Md. Saymon Ahammad, Jiangjiang Liu · 2025
This paper explores the evolution of computational methodologies in breast cancer (BC) detection, with a focus on recent advancements in artificial intelligence (AI)-driven systems. By analyzing state-of-the-art datasets and preprocessing techniques, we identify key factors that influence the success of AI-powered diagnostic tools. The study highlights the challenges faced by current approaches, including dataset limitations, class imbalance, and underrepresentation of certain imaging modalities. Furthermore, it underscores the need for advanced preprocessing strategies to enhance diagnostic accuracy. This research systematically examines studies published over the past six years (2018–2024), evaluates publicly available benchmark datasets, and discusses their strengths and weaknesses. Additionally, a range of data preprocessing techniques is explored, emphasizing their impact on breast cancer classification. The paper offers valuable insights into the challenges and future directions in the field, recommending the expansion of datasets, improvement of data standardization, and the integration of multimodal data for more robust AI-driven breast cancer diagnosis. Addressing the identified gaps in dataset diversity, imaging modality representation, and preprocessing methods will significantly advance the accuracy and real-world applicability of AI in breast cancer diagnostics.