Medical imaging generates vast amounts of visual and sequential data that clinicians must interpret quickly and accurately. Over the past decade deep learning (DL) has transformed this domain by enabling end-to-end learning of rich representations from raw images and signals. This chapter reviews the mathematical foundations, contemporary algorithms and practical applications of DL in healthcare. Real-world examples are drawn from radiology, pathology, ophthalmology, cardiology, dermatology and remote sensing. After introducing morphological image operations and evaluation metrics, traditional pipelines are contrasted with convolutional, transformer and graph-based architectures for detection, segmentation and recognition. Public datasets and standard evaluation protocols are summarised, followed by interdisciplinary applications and a small experimental analysis of dataset sizes, class imbalance and model-size trade-offs. Such challenges are the lack of data, interpretability and privacy which are addressed and the future research directions.