Journal: Journal of Imaging

Article Title: Retinal Disease Detection Using Deep Learning Techniques: A Comprehensive Review

doi: 10.3390/jimaging9040084

Figure Lengend Snippet: Summary of Deep Learning Methods for DR Classification.

Article Snippet: [ ] , ConvNet , EyePACS, e-optha, DiaretDB1 , , , , 0.954, 0.949, 0.955.

Techniques:

Journal: Nature Communications

Article Title: Training confounder-free deep learning models for medical applications

doi: 10.1038/s41467-020-19784-9

Figure Lengend Snippet: Balanced accuracy (%), precision (%), recall (%), and F 1 score of HIV-diagnosis prediction.

Article Snippet: We experimented on the 12,611 training images with ground-truth bone age (127.3 ± 41.2) and the ConvNet model publicly released on the Kaggle challenge page .

Techniques:

Journal: Nature Communications

Article Title: Training confounder-free deep learning models for medical applications

doi: 10.1038/s41467-020-19784-9

Figure Lengend Snippet: a Age discrepancy ( p = 0.0002, two-tailed two-sample t -test) between n = 223 control (Ctrl) subjects and n = 122 HIV patients resulted in the baseline ConvNet learning the confounding effects ( b , d , f ), which were alleviated by the proposed CF-Net ( c , e , g ). Boxplots are characterized by minimum, first quartile, median, third quartile, and maximum. b , c HIV-prediction scores measured on a subset of n = 122 control and n = 122 HIV subjects with the same age distribution ( c -independent). d , e t-SNE visualization of the feature space learned by the deep-learning models. f , g Saliency maps corresponding to the voxel-level attention (larger attention means more discriminative voxels) by the models.

Article Snippet: We experimented on the 12,611 training images with ground-truth bone age (127.3 ± 41.2) and the ConvNet model publicly released on the Kaggle challenge page .

Techniques: Two Tailed Test, Control

Journal: Nature Communications

Article Title: Training confounder-free deep learning models for medical applications

doi: 10.1038/s41467-020-19784-9

Figure Lengend Snippet: BAcc (precision and recall) on predicting sex from MRIs of NCANDA matched with respect to PDS. Optimal results were achieved when conditioning CF-Net on boys.

Article Snippet: We experimented on the 12,611 training images with ground-truth bone age (127.3 ± 41.2) and the ConvNet model publicly released on the Kaggle challenge page .

Techniques:

Journal: Nature Communications

Article Title: Training confounder-free deep learning models for medical applications

doi: 10.1038/s41467-020-19784-9

Figure Lengend Snippet: a Difference in the age distribution between n = 6, 833 boys and n = 5, 778 girls of the RSNA bone-age dataset ( p < 0.0001, two-tailed two-sample t -test). b Ground truth vs. predicted age of the ConvNet. ConvNet tended to predict higher age for girls than boys, indicating a confounding effect of sex. c This prediction gap between boys and girls was more pronounced in the age range of 110–200 months, but was significantly reduced by CF-Net, which modeled the dependency between F and c on a y -conditioned cohort. d Absolute prediction error (in months) of n = 3, 153 testing subjects produced by ConvNet and CF-Net with (or without) conditioning. Boxplots are characterized by minimum, first quartile, median, third quartile, and maximum. CF-Net with conditioning resulted in the most accurate prediction ( p < 0.0001, two-tailed two-sample t -test).

Article Snippet: We experimented on the 12,611 training images with ground-truth bone age (127.3 ± 41.2) and the ConvNet model publicly released on the Kaggle challenge page .

Techniques: Two Tailed Test, Produced

Journal: International Journal of Environmental Research and Public Health

Article Title: Epileptic Seizures Detection Using Deep Learning Techniques: A Review

doi: 10.3390/ijerph18115780

Figure Lengend Snippet: Summary of related works done using 1D-CNNs.

Article Snippet: [ ] , Deep ConvNet , 14 , Softmax , 80.00.

Techniques:

Journal: International Journal of Environmental Research and Public Health

Article Title: Epileptic Seizures Detection Using Deep Learning Techniques: A Review

doi: 10.3390/ijerph18115780

Figure Lengend Snippet: Summary of DL methods employed for automated detection of epileptic seizures.

Article Snippet: [ ] , Deep ConvNet , 14 , Softmax , 80.00.

Techniques: Extraction, Selection, Stripping Membranes

Journal: bioRxiv

Article Title: Cell type classification and unsupervised morphological phenotype identification from low-res images with deep learning

doi: 10.1101/533216

Figure Lengend Snippet: a. Bench-top microscope model [info] used for capturing cell images in commonly-used, thick bottomed, plastic cell culture flasks, b. Example brightfield images of both multi-cell regions (1216×1616 pixels) and manually cropped single cell regions (224×224 pixels). Note the camera used was not monochromatic and the images were converted to grayscale to reduce dimensionality and improve training efficiency, c. Quantile normalization of the image intensity distribution was performed for each cell to a reference distribution constructed from an arbitrarily selected single cell image, d. Cartoon of the proposed ConvNet architecture. Six quadruplets of convolutional, ReLU, Batch Normalization and average pooling layers were constructed. A single fully connected layer was constructed before the Softmax and classification layers, e. Illustration of Cell Type Classification from low-res flask images through artificial neural network. Labelled cells (yellow, blue) in labelled flasks should be fed into the neural network as training data, and unlabelled flasks shoule be treated as validation data. The trained ConvNet model is able to predict cell type with low error given novel cells with unknown label.

Article Snippet: A graphical representation of the Self-Label ConvNet designed for cell morphological phenotype clustering within one cell type via MATLAB 2018a (MathWorks) was displayed in .

Techniques: Microscopy, Cell Culture, Construct

Journal: bioRxiv

Article Title: Cell type classification and unsupervised morphological phenotype identification from low-res images with deep learning

doi: 10.1101/533216

Figure Lengend Snippet: a. Self-Label ConvNet Architecture Illustration. The group of augemented copies for each cell are considered unique classes, yielding the same number of classes in the final layer as there are cells used to train the network. The [l]ast [c]onvolutional [a]ctivation or ‘LCA’ feature space, labeled in green, is the structure of interest for the following mophological phenotype clustering, b. Training profile of Self-Label ConvNet. An accuracy of nearly 100% can be achieved for both training data and validation data, and a Softmax loss of nearly 0 can be achieved for both training data and validatioan data, c. Workflow for acquiring the LCA feature space an example cell. Novel cells are input into the pre-trained Self-Label ConvNet and the activations of the last convolutional layer are recorded as 32 3×3 matrices for each cell input. The matrices are then flattened to a vector of length 288, each element representing one ‘feature’ of the input cell, d. LCA matrix: LCA feature maps for many cells across all densities (2208 cells total) are displayed as rows in a matrix (size 2208×288) with each column representing one feature in the LCA, e. Clustering outcome for the LCA matrix applying k -means to rows according to Euclidean distance with k = 11. Clusters are shown after reshuffling the cell indices based on their cluster index., f. Cross-density cluster comparison. Two flasks of two densities are shown. For each flask, the fraction of cells belonging to each of the k = 11 clusters are displayed. Clusters with significantly different representations between densities are colored, g. Morphological Analysis: Two clusters of cells dominated by low density (cluster #10) and high density (cluster #3) respectively, were analyzed. Morphological properties for cells within these clusters were calculated with CellProfiler, and two features (SkeletonEndpoints and TextureSumVariance5) were chosen to generate a 2D projection, illustrating clear distinguishability in a low dimensional morphological feature space. High density biased cluster (cluster #3) was labeled in red and low density biased cluster (cluster #10) was labeled in blue.

Article Snippet: A graphical representation of the Self-Label ConvNet designed for cell morphological phenotype clustering within one cell type via MATLAB 2018a (MathWorks) was displayed in .

Techniques: Labeling, Plasmid Preparation

Journal: Sensors (Basel, Switzerland)

Article Title: Brain MRI Analysis for Alzheimer’s Disease Diagnosis Using CNN-Based Feature Extraction and Machine Learning

doi: 10.3390/s22082911

Figure Lengend Snippet: Summary and comparison of the selected recent research.

Article Snippet: Bäckström et al. (2018) [ ] , 3D ConvNet + Softmax , 96% , , .

Techniques: Comparison, Selection, T-Test, Generated

Journal: Sensors (Basel, Switzerland)

Article Title: Brain MRI Analysis for Alzheimer’s Disease Diagnosis Using CNN-Based Feature Extraction and Machine Learning

doi: 10.3390/s22082911

Figure Lengend Snippet: Comparison of our test performance with eight existing state-of-the-art methods.

Article Snippet: Bäckström et al. (2018) [ ] , 3D ConvNet + Softmax , 96% , , .

Techniques: Comparison