Explainable AI (XAI) methods are gaining prominence in medical imaging, addressing the critical need for transparency and trust in AI-driven diagnostic tools. Mammography, as the cornerstone of early breast cancer detection, holds immense potential for improving outcomes when integrated with AI solutions. However, widespread adoption of AI in clinical settings depends on explainability, which enhances clinicians' confidence in these tools. By exploring various XAI techniques and evaluating their strengths and weaknesses, researchers can significantly advance precision medicine. This review synthesizes existing research on XAI in medical imaging, focusing on mammography, a domain often overlooked in XAI studies. It provides a comparative analysis of XAI techniques employed in mammography, assessing their diagnostic efficacy and identifying research gaps, such as the lack of specialized evaluation frameworks. Additionally, the review examines evaluation methods for XAI in medical imaging and proposes modifications tailored to mammography diagnostics. Insights from XAI advancements in other fields are also explored for their potential to enhance interpretability and clinical relevance in breast cancer detection. The study concludes by highlighting critical research gaps and proposing directions for developing reliable, effective AI models that integrate XAI to transform breast cancer diagnostics.

Convolutional neural networks (CNNs) have previously been broadly utilized to binarize document images. These methods have problems when faced with degraded historical documents. This paper proposes the utilization of CNNs to identify foreground pixels using novel input-generated multichannel images. To create the images, the original source image is decomposed into wavelet subbands. Then, the original image is approximated by each subband separately, and finally, the multichannel image is constituted by arranging the original source image (grayscale image) as the first channel and the approximated image by each subband as the remaining channels. To achieve the best results, two scenarios are considered, that is, two-channel and four-channel images, and then fed into two types of CNN architectures, namely, single and multiple streams. To investigate the effect of the multichannel images proposed as network inputs, the CNNs used in the architectures are three popular networks, namely, U-net, SegNet, and DeepLabv3+. The experimental results of the scenarios demonstrate that our method is more successful than the three CNNs when trained by the original source images and proves competitive performance in comparison with state-of-the-art results using the DIBCO database.