A Path along Deep Learning for Medical Image Analysis : : With Focus on Burn Wounds and Brain Tumors.

• Superior document: Linköping Studies in Science and Technology. Dissertations Series ; v.2175
• Author: Cirillo, Marco Domenico.
• Place / Publishing House: Linköping : : Linkopings Universitet,, 2021.; {copy}2021.
• Year of Publication: 2021
• Edition: 1st ed.
• Language: English
• Series: Linköping Studies in Science and Technology. Dissertations Series
• Physical Description: 1 online resource (101 pages)

Table of Contents

Intro -- Abstract -- Acknowledgments -- Contents -- List of Figures -- Introduction -- Aim -- Delimitations -- Research questions -- Included papers -- Research ethics -- Outline -- Burn Wounds and Brain Tumors -- Burn wounds -- Pathophysiology -- Assessment methods -- Brain tumors -- Pathophysiology -- Assessment methods -- Reflections -- Image Features -- Type of features -- Color features -- Edge feature -- Texture features -- Mixed features -- Principal component analysis -- Independent component analysis -- Tensor decomposition -- Deep features -- Convolution -- Deep features -- Reflections -- Convolutional Neural Networks -- Deep learning basics -- Loss functions -- Forward and backward propagation -- Data pre-processing -- Weight initialization -- Normalization layers -- Activation functions -- Optimization -- Regularization -- Residual block -- Convolutional neural networks -- Convolutional layers -- CNNs for image classification -- CNNs for image segmentation -- CNNs for image generation -- Reflections -- Image Augmentation -- Image Augmentation Techniques -- Patch extraction -- Flipping -- Rotation -- Scaling -- Elastic grid-based deformation -- Brightness -- Reflections -- Generative Adversarial Networks -- Generator and discriminator -- GANs in medical imaging -- GAN losses -- Image-to-image GANs -- Pix2Pix -- Semantic image synthesis with spatially-adaptive normalization -- Reflections -- Papers, Discussions and Conclusions -- Paper I: Tensor decomposition for colour image segmentation of burn wounds -- Paper II: Time-independent prediction of burn depth using deep convolutional neural networks -- Paper III: Improving burn depth assessment for pediatric scalds by AI based on semantic segmentation of polarized light photography images -- Paper IV: Vox2Vox: 3D-GAN for brain tumour segmentation.

Paper V: What is the best data augmentation for 3D brain tumor segmentation? -- Conclusions -- Bibliography -- Papers.