Abstract We present in this paper the application of deep convolutional neural networks (CNNs), which is a state-of-the-art artificial intelligence (AI) approach machine learning, for automated time-independent prediction burn depth. Color images four types depth injured first few days, including normal skin and background, acquired by TiVi camera were trained tested with pretrained CNNs: VGG-16, GoogleNet, ResNet-50, ResNet-101. In end, best 10-fold cross-validation results obtained from...

Training segmentation networks requires large annotated datasets, which in medical imaging can be hard to obtain. Despite this fact, data augmentation has our opinion not been fully explored for brain tumor segmentation. In project we apply different types of (flipping, rotation, scaling, brightness adjustment, elastic deformation) when training a standard 3D U-Net, and demonstrate that significantly improves the network's performance many cases. Our conclusion is deformation work best,...

Abstract Research in burns has been a continuing demand over the past few decades, and important advancements are still needed to facilitate more effective patient stabilization reduce mortality rate. Burn wound assessment, which is an task for surgical management, largely depends on accuracy of burn area depth estimates. Automated quantification these parameters plays essential role reducing estimate errors conventionally carried out by clinicians. The automated calculation known as image...

This paper illustrates the efficacy of an artificial intelligence (AI) (a convolutional neural network, based on U-Net), for burn-depth assessment using semantic segmentation polarized high-performance light camera images burn wounds. The proposed method is evaluated paediatric scald injuries to differentiate four wound depths: superficial partial-thickness (healing in 0-7 days), intermediate 8-13 deep 14-20 after 21 days) and full-thickness burns, observed healing time. In total 100 were...

Gliomas are the most common primary brain malignancies, with different degrees of aggressiveness, variable prognosis and various heterogeneous histological sub-regions, i.e., peritumoral edema, necrotic core, enhancing non-enhancing tumour core. Although tumours can easily be detected using multi-modal MRI, accurate tumor segmentation is a challenging task. Hence, data provided by BraTS Challenge 2020, we propose 3D volume-to-volume Generative Adversarial Network for tumours. The model,...

Training segmentation networks requires large annotated datasets, which in medical imaging can be hard to obtain. Despite this fact, data augmentation has our opinion not been fully explored for brain tumor segmentation. In project we apply different types of (flipping, rotation, scaling, brightness adjustment, elastic deformation) when training a standard 3D U-Net, and demonstrate that significantly improves the network's performance many cases. Our conclusion is deformation work best,...