Deep Learning-Based Cattle Vocal Classification Model and Real-Time Livestock Monitoring System with Noise Filtering.
Ontology highlight
ABSTRACT:
SUBMITTER: Jung DH
PROVIDER: S-EPMC7911430 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Biological measurements are often contaminated with large amounts of non-stationary noise which require effective noise reduction techniques. We present a new real-time deep learning algorithm which produces adaptively a signal opposing the noise so that destructive interference occurs. As a proof of concept, we demonstrate the algorithm’s performance by reducing electromyogram noise in electroencephalograms with the usage of a custom, flexible, 3D-printed, compound electrode. With this setup, an average of 4dB and a maximum of 10dB improvement of the signal-to-noise ratio of the EEG was achieved by removing wide band muscle noise. This concept has the potential to not only adaptively improve the signal-to-noise ratio of EEG but can be applied to a wide range of biological, industrial and consumer applications such as industrial sensing or noise cancelling headphones.
| S-EPMC9678292 | biostudies-literature
Project description:Recent studies have shown that deep learning is capable of classifying dermatoscopic images at least as well as dermatologists. However, many studies in skin cancer classification utilize non-biopsy-verified training images. This imperfect ground truth introduces a systematic error, but the effects on classifier performance are currently unknown. Here, we systematically examine the effects of label noise by training and evaluating convolutional neural networks (CNN) with 804 images of melanoma and nevi labeled either by dermatologists or by biopsy. The CNNs are evaluated on a test set of 384 images by means of 4-fold cross validation comparing the outputs with either the corresponding dermatological or the biopsy-verified diagnosis. With identical ground truths of training and test labels, high accuracies with 75.03% (95% CI: 74.39-75.66%) for dermatological and 73.80% (95% CI: 73.10-74.51%) for biopsy-verified labels can be achieved. However, if the CNN is trained and tested with different ground truths, accuracy drops significantly to 64.53% (95% CI: 63.12-65.94%, p < 0.01) on a non-biopsy-verified and to 64.24% (95% CI: 62.66-65.83%, p < 0.01) on a biopsy-verified test set. In conclusion, deep learning methods for skin cancer classification are highly sensitive to label noise and future work should use biopsy-verified training images to mitigate this problem.
| S-EPMC7218064 | biostudies-literature
Project description:Application of scanning transmission electron microscopy (STEM) to in situ observation will be essential in the current and emerging data-driven materials science by taking STEM's high affinity with various analytical options into account. As is well known, STEM's image acquisition time needs to be further shortened to capture a targeted phenomenon in real-time as STEM's current temporal resolution is far below the conventional TEM's. However, rapid image acquisition in the millisecond per frame or faster generally causes image distortion, poor electron signals, and unidirectional blurring, which are obstacles for realizing video-rate STEM observation. Here we show an image correction framework integrating deep learning (DL)-based denoising and image distortion correction schemes optimized for STEM rapid image acquisition. By comparing a series of distortion corrected rapid scan images with corresponding regular scan speed images, the trained DL network is shown to remove not only the statistical noise but also the unidirectional blurring. This result demonstrates that rapid as well as high-quality image acquisition by STEM without hardware modification can be established by the DL. The DL-based noise filter could be applied to in-situ observation, such as dislocation activities under external stimuli, with high spatio-temporal resolution.
| S-EPMC9356044 | biostudies-literature
Project description:Point-of-care (POC) real-time polymerase chain reaction (PCR) has become one of the most important technologies for many fields such as pathogen detection and water-quality monitoring. POC real-time PCR usually adopts chips with small-volume chambers for portability, which is more likely to produce complex noise that seriously affects the accuracy. Such complex noises are difficult to be eliminated by the traditional fixed area algorithm that is most commonly used at present because they usually have random shape, location, and brightness. To address this problem, we proposed a novel image analysis method, Dynamic Deep Learning Noise Elimination Method (DIPLOID), in this paper. Our new method could recognize and output the mask of the interference by Mask R-CNN, and then subtract the interference and select the maximum valid contiguous area for brightness analysis by dynamic programming. Compared with the traditional method, DIPLOID increased the accuracy, sensitivity, and specificity from 57.9 to 94.6%, 49.1 to 93.9%, and 65.9 to 95.2%, respectively. DIPLOID has great anti-interference, robustness, and sensitivity, which can reduce the impact of complex noise as much as possible from the aspect of the algorithm. As shown in the experiments of this paper, our method significantly improved the accuracy to over 94% under the complex noise situation, which could make the POC real-time PCR have greater potential in the future.
| S-EPMC8976108 | biostudies-literature
Project description:Accidents related to the administration of intravenous (IV) medication, such as drug overdose/underdose, drug/patient mis-identification, and delayed bag exchange, occur consistently in clinical fields. Several previous studies have suggested various contact-sensing and image-processing methodologies; however, most of them can increase the workload of nursing staffs during the long-term, continuous monitoring. In this study, we proposed a smart IV pole that can monitor the infusion status of up to four IV medications (patient/drug identification, and liquid residue) with various sizes and hanging positions to reduce IV-related accidents and improve patient safety with the least additional workload; the system consists of 12 cameras, one code scanner, and four controllers. Two types of deep learning models for automated camera selection (CNN-1) and liquid residue monitoring (CNN-2), and three drug residue estimation equations were implemented. The experimental results demonstrated that the accuracy of identification code-checking (60 tests) was 100%. The classification accuracy and the mean inference time of CNN-1 (1200 tests) were 100% and 140 ms. The mean average precision and the mean inference time of CNN-2 (300 tests) were 0.94 and 144 ms. The average error rates between the alarm setting (20, 30, and 40 mL) and the actual drug residue when the alarm first generated were 4.00%, 7.33%, and 4.50% for a 1,000 mL bag; 6.00%, 4.67%, and 2.50% for a 500 mL bag; and 3.00%, 6.00%, and 3.50% for a 100 mL bag, respectively. Our results suggest that the implemented AI-based prototype IV pole is a potential tool for reducing IV-related accidents and improving in-hospital patient safety.Supplementary informationThe online version contains supplementary material available at 10.1007/s13534-023-00292-w.
| S-EPMC10245361 | biostudies-literature
Project description:ObjectiveTo address the challenge of assessing sedation status in critically ill patients in the intensive care unit (ICU), we aimed to develop a non-contact automatic classifier of agitation using artificial intelligence and deep learning.MethodsWe collected the video recordings of ICU patients and cut them into 30-second (30-s) and 2-second (2-s) segments. All of the segments were annotated with the status of agitation as "Attention" and "Non-attention". After transforming the video segments into movement quantification, we constructed the models of agitation classifiers with Threshold, Random Forest, and LSTM and evaluated their performances.ResultsThe video recording segmentation yielded 427 30-s and 6405 2-s segments from 61 patients for model construction. The LSTM model achieved remarkable accuracy (ACC 0.92, AUC 0.91), outperforming other methods.ConclusionOur study proposes an advanced monitoring system combining LSTM and image processing to ensure mild patient sedation in ICU care. LSTM proves to be the optimal choice for accurate monitoring. Future efforts should prioritize expanding data collection and enhancing system integration for practical application.
| S-EPMC10946151 | biostudies-literature
Project description:In this study, a simple and customizable convolution neural network framework was used to train a vibration classification model that can be integrated into the measurement application in order to realize accurate and real-time bridge vibration status on mobile platforms. The inputs for the network model are basically the multichannel time-series signals acquired from the built-in accelerometer sensor of smartphones, while the outputs are the predefined vibration categories. To verify the effectiveness of the proposed framework, data collected from long-term monitoring of bridge were used for training a model, and its classification performance was evaluated on the test set constituting the data collected from the same bridge but not used previously for training. An iOS application program was developed on the smartphone for incorporating the trained model with predefined classification labels so that it can classify vibration datasets measured on any other bridges in real-time. The results justify the practical feasibility of using a low-latency, high-accuracy smartphone-based system amid which bottlenecks of processing large amounts of data will be eliminated, and stable observation of structural conditions can be promoted.
| S-EPMC7248740 | biostudies-literature
Project description:Anthropogenic noise imposes novel selection pressures, especially on species that communicate acoustically. Many animals-including insects, frogs, whales and birds-produce sounds at higher frequencies in areas with low-frequency noise pollution. Although there is support for animals changing their vocalizations in real time in response to noise (i.e. immediate flexibility), other evolutionary mechanisms for animals that learn their vocalizations remain largely unexplored. We hypothesize that cultural selection for signal structures less masked by noise is a mechanism of acoustic adaptation to anthropogenic noise. We test this hypothesis by presenting nestling white-crowned sparrows (Zonotrichia leucophyrs) with less-masked (higher-frequency) and more-masked (lower-frequency) tutor songs either during playback of anthropogenic noise (noise-tutored treatment) or at a different time from noise playback (control treatment). As predicted, we find that noise-tutored males learn less-masked songs significantly more often, whereas control males show no copying preference, providing strong experimental support for cultural selection in response to anthropogenic noise. Further, noise-tutored males reproduce songs at higher frequencies than their tutor, indicating a distinct mechanism to increase signal transmission in a noisy environment. Notably, noise-tutored males achieve lower performance songs than their tutors, suggesting potential costs in a sexual selection framework.
| S-EPMC6191693 | biostudies-literature
Project description:Here we present miR-eCLIP analysis of AGO2 in HEK293 cells to address the small RNA repertoire and uncover their physiological targets. We developed an optimized bioinformatics approach of chimeric read identification to detect chimeras of high confidence, which were useed as an biologically validated input for miRBind, a deep learning method and web-server that can be used to accurately predict the potential of miRNA:target site binding.
2022-12-22 | GSE218466 | GEO
Project description:Background and objectiveMagnetic resonance (MR) imaging is increasingly used in studies of speech as it enables non-invasive visualisation of the vocal tract and articulators, thus providing information about their shape, size, motion and position. Extraction of this information for quantitative analysis is achieved using segmentation. Methods have been developed to segment the vocal tract, however, none of these also fully segment any articulators. The objective of this work was to develop a method to fully segment multiple groups of articulators as well as the vocal tract in two-dimensional MR images of speech, thus overcoming the limitations of existing methods.MethodsFive speech MR image sets (392 MR images in total), each of a different healthy adult volunteer, were used in this work. A fully convolutional network with an architecture similar to the original U-Net was developed to segment the following six regions in the image sets: the head, soft palate, jaw, tongue, vocal tract and tooth space. A five-fold cross-validation was performed to investigate the segmentation accuracy and generalisability of the network. The segmentation accuracy was assessed using standard overlap-based metrics (Dice coefficient and general Hausdorff distance) and a novel clinically relevant metric based on velopharyngeal closure.ResultsThe segmentations created by the method had a median Dice coefficient of 0.92 and a median general Hausdorff distance of 5mm. The method segmented the head most accurately (median Dice coefficient of 0.99), and the soft palate and tooth space least accurately (median Dice coefficients of 0.92 and 0.93 respectively). The segmentations created by the method correctly showed 90% (27 out of 30) of the velopharyngeal closures in the MR image sets.ConclusionsAn automatic method to fully segment multiple groups of articulators as well as the vocal tract in two-dimensional MR images of speech was successfully developed. The method is intended for use in clinical and non-clinical speech studies which involve quantitative analysis of the shape, size, motion and position of the vocal tract and articulators. In addition, a novel clinically relevant metric for assessing the accuracy of vocal tract and articulator segmentation methods was developed.
| S-EPMC7732702 | biostudies-literature