The nematode Caenorhabditis elegans (C. elegans) is often used as an alternative animal model due to several advantages such as morphological changes that can be seen directly under a microscope. Limitations of the model include the usage of expensive and cumbersome microscopes, and restrictions of the comprehensive use of C. elegans for toxicological trials. With the general applicability of the detection of C. elegans from microscope images via machine learning, as well as of smartphone-based microscopes, this article investigates the suitability of smartphone-based microscopy to detect C. elegans in a complete Petri dish. Thereby, the article introduces a smartphone-based microscope (including optics, lighting, and housing) for monitoring C. elegans and the corresponding classification via a trained Histogram of Oriented Gradients (HOG) feature-based Support Vector Machine for the automatic detection of C. elegans. Evaluation showed classification sensitivity of 0.90 and specificity of 0.85, and thereby confirms the general practicability of the chosen approach.

Caenorhabditis elegans (C. elegans) is an important model organism for studying molecular genetics, developmental biology, neuroscience, and cell biology. Advantages of the model organism include its rapid development and aging, easy cultivation, and genetic tractability. C. elegans has been proven to be a well-suited model to study toxicity with identified toxic compounds closely matching those observed in mammals. For phenotypic screening, especially the worm number and the locomotion are of central importance. Traditional methods such as human counting or analyzing high-resolution microscope images are time-consuming and rather low throughput. The article explores the feasibility of low-cost, low-resolution do-it-yourself microscopes for image acquisition and automated evaluation by deep learning methods to reduce cost and allow high-throughput screening strategies. An image acquisition system is proposed within these constraints and used to create a large data-set of whole Petri dishes containing C. elegans. By utilizing the object detection framework Mask R-CNN, the nematodes are located, classified, and their contours predicted. The system has a precision of 0.96 and a recall of 0.956, resulting in an F1-Score of 0.958. Considering only correctly located C. elegans with an AP@0.5 IoU, the system achieved an average precision of 0.902 and a corresponding F1 Score of 0.906.

In this paper, we investigate the viability of a variational U-Net architecture for denoising of single-channel audio data. Deep network speech enhancement systems commonly aim to estimate filter masks, or opt to skip preprocessing steps to directly work on the waveform signal, potentially neglecting relationships across higher dimensional spectro-temporal features. We study the adoption of a probabilistic bottleneck, as well as dilated convolutions, into the classic U-Net architecture. Evaluation of a number of network variants is carried out using signal-to-distortion ratio and perceptual model scores, with audio data including known and unknown noise types as well as reverberation. Our experiments show that the residual (skip) connections in the proposed system are required for successful end-to-end signal enhancement, i.e., without filter mask estimation. Further, they indicate a slight advantage of the variational U-Net architecture over its non-variational version in terms of signal enhancement performance under reverberant conditions. Specifically, PESQ scores show increases of 0.28 and 0.49 in reverberant and non-reverberant scenes, respectively. Anecdotal evidence points to improved suppression of impulsive noise sources with the variational end-to-end U-Net compared to the recurrent mask estimation network baseline.

In this paper, we propose to extend the deep, complex U-Network architecture for speech enhancement by incorporating a probabilistic (i.e., variational) latent space model. The proposed model is evaluated against several ablated versions of itself in order to study the effects of the variational latent space model, complex-value processing, and self-attention. Evaluation on the MS-DNS 2020 and Voicebank+Demand datasets yields consistently high performance. E.g., the proposed model achieves an SI-SDR of up to 20.2 dB, about 0.5 to 1.4 dB higher than its ablated version without probabilistic latent space, 2-2.4 dB higher than WaveUNet, and 6.7 dB above PHASEN. Compared to real-valued magnitude spectrogram processing with a variational U-Net, the complex U-Net achieves an improvement of up to 4.5 dB SI-SDR. Complex spectrum encoding as magnitude and phase yields best performance in anechoic conditions whereas real and imaginary part representation results in better generalization to (novel) reverberation conditions, possibly due to the underlying physics of sound.

No abstract