The growing incidence of skin cancer makes computer-aided diagnosis tools for this group of diseases increasingly important. The use of ultrasound has the potential to complement information from optical dermoscopy. The current work presents a fully automatic classification framework utilizing fully-automated (FA) segmentation and compares it with classification using two semi-automated (SA) segmentation methods. Ultrasound recordings were taken from a total of 310 lesions (70 melanoma, 130 basal cell carcinoma and 110 benign nevi). A support vector machine (SVM) model was trained on 62 features, with ten-fold cross-validation. Six classification tasks were considered, namely all the possible permutations of one class versus one or two remaining classes. The receiver operating characteristic (ROC) area under the curve (AUC) as well as the accuracy (ACC) were measured. The best classification was obtained for the classification of nevi from cancerous lesions (melanoma, basal cell carcinoma), with AUCs of over 90% and ACCs of over 85% obtained with all segmentation methods. Previous works have either not implemented FA ultrasound-based skin cancer classification (making diagnosis more lengthy and operator-dependent), or are unclear in their classification results. Furthermore, the current work is the first to assess the effect of implementing FA instead of SA classification, with FA classification never degrading performance (in terms of AUC or ACC) by more than 5%.

Psoriasis is a chronic inflammatory disease with unmet medical needs. To clarify potential therapeutic targets, different animal models have been developed. In the current study, imiquimod-induced psoriasiform dermatitis was used for monitoring the changes in skin thickness, transepidermal water loss, body weight, blood perfusion and drug permeability for a topical cream formulation of caffeine, both in wild type and in knock out mice. Morphological characterization of control and diseased tissues was performed by scanning electron microscopy and two-photon microscopy. The chemically induced psoriatic group showed increased skin permeability for the model drug during disease progression. In wild type and TRPA1 KO mice, however, enhanced skin thickness and hyperkeratosis blocked further increase of drug penetration at the late phase (96 h). These results indicate that topical drug therapy can be more effective in early phases of plaque development, when skin thickness is lower. Although paracellular connections (tight junctions) are looser in the advanced phase, hyperkeratosis blocks drug delivery through the transappendageal routes. Novel drug formulations may have the potency for effective drug delivery across the epidermal barrier even in the advanced phase. For development of more effective topical drugs, further research is proposed to explore drug penetration both in healthy and diseased conditions.

Genetic Algorithms are stochastic optimization methods where solution candidates, complying to a specific problem representation, are evaluated according to a predefined fitness function. These approaches can provide solutions in various tasks even, where analytic solutions can not be or are too complex to be computed. In this paper we will show, how certain set of problems are partially solvable allowing us to grade segments of a solution individually, which results local and individual tuning of mutation parameters for genes. We will demonstrate the efficiency of our method on the N-Queens and travelling salesman problems where we can demonstrate that our approach always results faster convergence and in most cases a lower error than the traditional approach.

Convolutional networks are commonly used in various machine learning tasks, and they are more and more popularly used in the embedded domain with devices such as smart cameras and mobile phones. The operation of convolution can be substituted by point-wise multiplication in the Fourier domain, which can save operation, but usually, it is applied with a Fourier transform before and an inverse Fourier transform after the multiplication, since other operations in neural networks cannot be implemented efficiently in the Fourier domain. In this paper, we will present a method for implementing neural network completely in the Fourier domain, and by this, saving multiplications and the operations of inverse Fourier transformations. Our method can decrease the number of operations by four times the number of pixels in the convolutional kernel with only a minor decrease in accuracy, for example, 4% on the MNIST and 2% on the HADB datasets.

<p>Earthquake detection and phase picking are central problems of seismic activity analysis. Traditional approaches [1] and machine learning methods [2] are applied in this domain, typically performing well on commonly investigated standard datasets reaching above 99% accuracy in seismic activity detection.</p><p> </p><p>Unfortunately, most databases in the literature contain only earthquake data as detectable activities and spurious activities such as mining are not included in these datasets. We have investigated a recently published deep neural network-based method [3] and found that these detectors are fooled by mining activity.</p><p> </p><p>To solve this problem, we have created a complex dataset that contains 1200 independently recorded mining and earthquake activities from Central Europe. Our dataset poses a more complex problem than commonly investigated datasets such as the STanford EArthquake Dataset and can be viewed as an extension of that.</p><p> </p><p>We have trained a convolutional neural network containing five convolutional and three fully-connected layers to classify these signals on this dataset and reached a 94% classification accuracy, which demonstrates that the categorization of mining activity and earthquakes is possible with modern machine learning approaches.</p><p><br><br></p><p>[1] Galiana-Merino, J. J., Rosa-Herranz, J. L., & Parolai, S. (2008). Seismic P Phase Picking Using a Kurtosis-Based Criterion in the Stationary Wavelet Domain. IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3815-3826.</p><p> </p><p>[2] Zhu, W., & Beroza, G. C. (2019). PhaseNet: a deep-neural-network-based seismic arrival-time picking method. Geophysical Journal International, 216(1), 261-273.</p><p> </p><p>[3] Mousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y., & Beroza, G. C. (2020). Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking. Nature communications, 11(1), 1-12.</p>