Journal: Scientific Reports

Article Title: UNI-EM: An Environment for Deep Neural Network-Based Automated Segmentation of Neuronal Electron Microscopic Images

doi: 10.1038/s41598-019-55431-0

Figure Lengend Snippet: Example workflow 1: Mitochondria segmentation using 2D CNN. ( A ) Conventional workflow. Users first paint the regions of mitochondria of a target EM image using painting software, e.g., VAST lite (1, top) . This mitochondrial segmentation image (ground truth) and the EM image are transferred to Tensorflow/Python for CNN training and inference (2,3; right). Inferred segmentation is then postprocessed (4, left), e.g., using imageJ, proofread and visualized by VAST lite (5, top). Such relays between software packages are necessary. ( B ) UNI-EM dropdown menu. A series of software (a-d) is located for the CNN-based segmentation (1–5). Standard png/tiff file format is used to connect these software packages. ( C ) Workflow in UNI-EM. Extended Dojo supports paint functions (1; top, left) to draw mitochondrial segmentation (top, right). Users can conduct CNN training (2) and inference (3) through a control panel. A labeling function is also implemented for postprocessing (4, each label is denoted by color). These segmented images are proofread by Dojo (5, left), and visualized by the 3D annotator (5, right).

Article Snippet: The segmentation accuracy of the 3D CNN was quantified as Jaccard 0.92, Dice 0.96, and conformity 0.91 (semantic segmentation; ATUM/SEM data), whereas that of our standard 2D CNN was quantified as Jaccard 0.91, Dice 0.95, conformity 0.90 (semantic segmentation).

Techniques: Software, Control, Labeling

Journal: Scientific Reports

Article Title: UNI-EM: An Environment for Deep Neural Network-Based Automated Segmentation of Neuronal Electron Microscopic Images

doi: 10.1038/s41598-019-55431-0

Figure Lengend Snippet: Underlying architecture of UNI-EM. UNI-EM has a heterogenous system. Present desktop computers have two types of computational resources: CPU and GPU (top). A GPU is used by Tensorflow for CNN computing (middle), which is not appropriate for shared use. Only the resource monitor Tensorboard can be used by remote users (bottom). Similarly, remote users can use proofreader Dojo and 3D annotator. Only a desktop user (silhouette person) can control all of the UNI-EM functions, including job submission for CNN computing such as training and inference.

Article Snippet: The segmentation accuracy of the 3D CNN was quantified as Jaccard 0.92, Dice 0.96, and conformity 0.91 (semantic segmentation; ATUM/SEM data), whereas that of our standard 2D CNN was quantified as Jaccard 0.91, Dice 0.95, conformity 0.90 (semantic segmentation).

Techniques: Control

Journal: Advances in Radiation Oncology

Article Title: Review of Deep Learning Based Autosegmentation for Clinical Target Volume: Current Status and Future Directions

doi: 10.1016/j.adro.2024.101470

Figure Lengend Snippet: The general characteristics and performance of DLAS model from each article included in this review

Article Snippet: Buelens et al, 2022 , Breast (CT) , International , Yes , 3D CNN vs manual , Yes , 95 , CNN segmentation performance was best for breast CTV and worse for Rotter's space and the internal mammary nodes Guideline consistency improved from 77.14%-90.71% in favor of CNN segmentation , CNN segmentation saved on average 24 min per patient with a median time of 35 min for pure manual segmentation , Not reported.

Techniques: Modification, Selection

Journal: Clinical Psychopharmacology and Neuroscience

Article Title: An Overview of Deep Learning Algorithms and Their Applications in Neuropsychiatry

doi: 10.9758/cpn.2021.19.2.206

Figure Lengend Snippet: Convolutional neural network (CNN). A CNN contains two basic parts: feature extraction and classification. The feature extraction part consists of successive convolutional and pooling layers. A convolutional layer applies convolutional filters called a kernel to the image for exploring low and high-level structures. These structures are obtained by shifting these kernels, so called convolution, in the image with a set of weights. After multiplying the elements of these kernels with the corresponding receiving field elements, a feature map is obtained. These maps are passed through nonlinear activation function (e.g., a rectified linear unit). The task of pooling layer is to reduce the feature map size and the total number of parameters to be optimized in the network. It works by gathering similar information in the neighborhood of the receptive field and find a representative value (e.g., maximum or average) within this local region. Flatten layer converts matrices from the convolution layers into a one-dimensional array for the next layer. Fully connected layer computes the final outputs using back propagation and gradient descent as for standard artificial neural networks.

Article Snippet: Korolev et al . 2017 [ ] , Plain 3D CNN (VoxCNN) and ResNet with six VoxRes blocks , 2017 , 3D sMRI , Classification of AD, EMCI, and LMCI , Accuray (VoxCNN) = 79% for NC vs. AD Accuray (VoxCNN) = 63% for NC vs. LMCI Accuray (ResNet) = 54% for AD vs. EMCI Accuray (ResNet) = 80% for NC vs. AD Accuray (ResNet) = 61% for NC vs. LMCI Accuray (ResNet) = 56% for AD vs. EMCI.

Techniques: Extraction, Activation Assay

Journal: Clinical Psychopharmacology and Neuroscience

Article Title: An Overview of Deep Learning Algorithms and Their Applications in Neuropsychiatry

doi: 10.9758/cpn.2021.19.2.206

Figure Lengend Snippet: Flow diagram for study selection (modified from Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement). ANNs, artificial neural networks; CNNs, convolutional neural networks; RNNs, recurrent neural networks; GANs, generative adversarial networks.

Article Snippet: Korolev et al . 2017 [ ] , Plain 3D CNN (VoxCNN) and ResNet with six VoxRes blocks , 2017 , 3D sMRI , Classification of AD, EMCI, and LMCI , Accuray (VoxCNN) = 79% for NC vs. AD Accuray (VoxCNN) = 63% for NC vs. LMCI Accuray (ResNet) = 54% for AD vs. EMCI Accuray (ResNet) = 80% for NC vs. AD Accuray (ResNet) = 61% for NC vs. LMCI Accuray (ResNet) = 56% for AD vs. EMCI.

Techniques: Selection, Modification

Journal: Clinical Psychopharmacology and Neuroscience

Article Title: An Overview of Deep Learning Algorithms and Their Applications in Neuropsychiatry

doi: 10.9758/cpn.2021.19.2.206

Figure Lengend Snippet: Studies using CNN in neuropsychiatry

Article Snippet: Korolev et al . 2017 [ ] , Plain 3D CNN (VoxCNN) and ResNet with six VoxRes blocks , 2017 , 3D sMRI , Classification of AD, EMCI, and LMCI , Accuray (VoxCNN) = 79% for NC vs. AD Accuray (VoxCNN) = 63% for NC vs. LMCI Accuray (ResNet) = 54% for AD vs. EMCI Accuray (ResNet) = 80% for NC vs. AD Accuray (ResNet) = 61% for NC vs. LMCI Accuray (ResNet) = 56% for AD vs. EMCI.

Techniques: Single Photon Emission Computed Tomography