This paper provides a comprehensive survey of deep learning applications in medical image analysis, focusing on over 300 recent contributions. It reviews major deep learning concepts and their applications in tasks such as image classification, object detection, segmentation, and registration. The survey covers various imaging modalities, including neuroimaging, retinal imaging, pulmonary imaging, digital pathology, breast imaging, cardiac imaging, abdominal imaging, and musculoskeletal imaging. The paper highlights the shift from handcrafted features to learned features using deep learning models, particularly convolutional neural networks (CNNs). It discusses the evolution of CNN architectures, from shallow models like LeNet and AlexNet to deeper models like VGG, Inception, and ResNet. The survey also explores multi-stream architectures for multi-scale and 2.5D image analysis, as well as segmentation techniques using fully convolutional networks (FCNs) and U-net. Additionally, it covers recurrent neural networks (RNNs) for sequence analysis and unsupervised models like auto-encoders and deep belief networks. The paper concludes with a critical discussion of open challenges and future research directions in deep learning for medical image analysis.This paper provides a comprehensive survey of deep learning applications in medical image analysis, focusing on over 300 recent contributions. It reviews major deep learning concepts and their applications in tasks such as image classification, object detection, segmentation, and registration. The survey covers various imaging modalities, including neuroimaging, retinal imaging, pulmonary imaging, digital pathology, breast imaging, cardiac imaging, abdominal imaging, and musculoskeletal imaging. The paper highlights the shift from handcrafted features to learned features using deep learning models, particularly convolutional neural networks (CNNs). It discusses the evolution of CNN architectures, from shallow models like LeNet and AlexNet to deeper models like VGG, Inception, and ResNet. The survey also explores multi-stream architectures for multi-scale and 2.5D image analysis, as well as segmentation techniques using fully convolutional networks (FCNs) and U-net. Additionally, it covers recurrent neural networks (RNNs) for sequence analysis and unsupervised models like auto-encoders and deep belief networks. The paper concludes with a critical discussion of open challenges and future research directions in deep learning for medical image analysis.