Machine learning based dynamic consensus model for predicting blood-brain barrier permeability

The blood-brain barrier (BBB) is an important defence mechanism that restricts disease-causing pathogens and toxins to enter the brain from the bloodstream. In recent years, many in silico methods were proposed for predicting BBB permeability, however, the reliability of these models is questionable due to the smaller and class-imbalance dataset which subsequently leads to a very high false positive rate. In this study, machine learning and deep learning-based predictive models were built using XGboost, Random Forest, Extra-tree classifiers and deep neural network. A dataset of 8153 compounds comprising both the BBB permeable and BBB non-permeable was curated and subjected to calculations of molecular descriptors and fingerprints for generating the features for machine learning and deep learning models. Three balancing techniques were then applied to the dataset to address the class-imbalance issue. A comprehensive comparison among the models showed that the deep neural network model generated on the balanced MACCS fingerprint dataset outperformed with an accuracy of 97.8% and a ROC-AUC score of 0.98 among all the models. Additionally, a dynamic consensus model was prepared with the machine learning models and validated with a benchmark dataset for predicting BBB permeability with higher confidence scores.

An early detection and segmentation of Brain Tumor using Deep Neural Network

BACKGROUND

Magnetic resonance image (MRI) brain tumor segmentation is crucial and important in the medical field, which can help in diagnosis and prognosis, overall growth predictions, Tumor density measures, and care plans needed for patients. The difficulty in segmenting brain Tumors is primarily because of the wide range of structures, shapes, frequency, position, and visual appeal of Tumors, like intensity, contrast, and visual variation. With recent advancements in Deep Neural Networks (DNN) for image classification tasks, intelligent medical image segmentation is an exciting direction for Brain Tumor research. DNN requires a lot of time & processing capabilities to train because of only some gradient diffusion difficulty and its complication.

METHODS

To overcome the gradient issue of DNN, this research work provides an efficient method for brain Tumor segmentation based on the Improved Residual Network (ResNet). Existing ResNet can be improved by maintaining the details of all the available connection links or by improving projection shortcuts. These details are fed to later phases, due to which improved ResNet achieves higher precision and can speed up the learning process.

RESULTS

The proposed improved Resnet address all three main components of existing ResNet: the flow of information through the network layers, the residual building block, and the projection shortcut. This approach minimizes computational costs and speeds up the process.

CONCLUSION

An experimental analysis of the BRATS 2020 MRI sample data reveals that the proposed methodology achieves competitive performance over the traditional methods like CNN and Fully Convolution Neural Network (FCN) in more than 10% improved accuracy, recall, and f-measure.

A multi-view co-training network for semi-supervised medical image-based prognostic prediction

Prognostic prediction has long been a hotspot in disease analysis and management, and the development of image-based prognostic prediction models has significant clinical implications for current personalized treatment strategies. The main challenge in prognostic prediction is to model a regression problem based on censored observations, and semi-supervised learning has the potential to play an important role in improving the utilization efficiency of censored data. However, there are yet few effective semi-supervised paradigms to be applied. In this paper, we propose a semi-supervised co-training deep neural network incorporating a support vector regression layer for survival time estimation (Co-DeepSVS) that improves the efficiency in utilizing censored data for prognostic prediction. First, we introduce a support vector regression layer in deep neural networks to deal with censored data and directly predict survival time, and more importantly to calculate the labeling confidence of each case. Then, we apply a semi-supervised multi-view co-training framework to achieve accurate prognostic prediction, where labeling confidence estimation with prior knowledge of pseudo time is conducted for each view. Experimental results demonstrate that the proposed Co-DeepSVS has a promising prognostic ability and surpasses most widely used methods on a multi-phase CT dataset. Besides, the introduction of SVR layer makes the model more robust in the presence of follow-up bias.

Automated semantic lung segmentation in chest CT images using deep neural network

Lung segmentation algorithms play a significant role in segmenting theinfected regions in the lungs. This work aims to develop a computationally efficient and robust deep learning model for lung segmentation using chest computed tomography (CT) images with DeepLabV3 + networks for two-class (background and lung field) and four-class (ground-glass opacities, background, consolidation, and lung field). In this work, we investigate the performance of the DeepLabV3 + network with five pretrained networks: Xception, ResNet-18, Inception-ResNet-v2, MobileNet-v2 and ResNet-50. A publicly available database for COVID-19 that contains 750 chest CT images and corresponding pixel-labeled images are used to develop the deep learning model. The segmentation performance has been assessed using five performance measures: Intersection of Union (IoU), Weighted IoU, Balance F1 score, pixel accu-racy, and global accuracy. The experimental results of this work confirm that the DeepLabV3 + network with ResNet-18 and a batch size of 8 have a higher performance for two-class segmentation. DeepLabV3 + network coupled with ResNet-50 and a batch size of 16 yielded better results for four-class segmentation compared to other pretrained networks. Besides, the ResNet with a fewer number of layers is highly adequate for developing a more robust lung segmentation network with lesser computational complexity compared to the conventional DeepLabV3 + network with Xception. This present work proposes a unified DeepLabV3 + network to delineate the two and four different regions automatically using CT images for CoVID-19 patients. Our developed automated segmented model can be further developed to be used as a clinical diagnosis system for CoVID-19 as well as assist clinicians in providing an accurate second opinion CoVID-19 diagnosis.

Prediction of KRASG12C inhibitors using conjoint fingerprint and machine learning-based QSAR models

Kirsten rat sarcoma virus G12C (KRAS^G12C) is the major protein mutation associated with non-small cell lung cancer (NSCLC) severity. Inhibiting KRAS^G12C is therefore one of the key therapeutic strategies for NSCLC patients. In this paper, a cost-effective data driven drug design employing machine learning-based quantitative structure-activity relationship (QSAR) analysis was built for predicting ligand affinities against KRAS^G12C protein. A curated and non-redundant dataset of 1033 compounds with KRAS^G12C inhibitory activity (pIC₅₀) was used to build and test the models. The PubChem fingerprint, Substructure fingerprint, Substructure fingerprint count, and the conjoint fingerprint-a combination of PubChem fingerprint and Substructure fingerprint count-were used to train the models. Using comprehensive validation methods and various machine learning algorithms, the results clearly showed that the XGBoost regression (XGBoost) achieved the highest performance in term of goodness of fit, predictivity, generalizability and model robustness (R² = 0.81, Q²_CV = 0.60, Q²_Ext = 0.62, R² - Q²_Ext = 0.19, R²_Y-Random = 0.31 ± 0.03, Q²_Y-Random = -0.09 ± 0.04). The top 13 molecular fingerprints that correlated with the predicted pIC₅₀ values were SubFPC274 (aromatic atoms), SubFPC307 (number of chiral-centers), PubChemFP37 (≥1 Chlorine), SubFPC18 (Number of alkylarylethers), SubFPC1 (number of primary carbons), SubFPC300 (number of 1,3-tautomerizables), PubChemFP621 (N-C:C:C:N structure), PubChemFP23 (≥1 Fluorine), SubFPC2 (number of secondary carbons), SubFPC295 (number of C-ONS bonds), PubChemFP199 (≥4 6-membered rings), PubChemFP180 (≥1 nitrogen-containing 6-membered ring), and SubFPC180 (number of tertiary amine). These molecular fingerprints were virtualized and validated using molecular docking experiments. In conclusion, this conjoint fingerprint and XGBoost-QSAR model demonstrated to be useful as a high-throughput screening tool for KRAS^G12C inhibitor identification and drug design.

A deep neural network framework to derive interpretable decision rules for accurate traumatic brain injury identification of infants

OBJECTIVE

We aimed to develop a robust framework to model the complex association between clinical features and traumatic brain injury (TBI) risk in children under age two, and identify significant features to derive clinical decision rules for triage decisions.

METHODS

In this retrospective study, four frequently used machine learning models, i.e., support vector machine (SVM), random forest (RF), deep neural network (DNN), and XGBoost (XGB), were compared to identify significant clinical features from 24 input features associated with the TBI risk in children under age two under the permutation feature importance test (PermFIT) framework by using the publicly available data set from the Pediatric Emergency Care Applied Research Network (PECARN) study. The prediction accuracy was determined by comparing the predicted TBI status with the computed tomography (CT) scan results since CT scan is the gold standard for diagnosing TBI.

RESULTS

At a significance level of [Formula: see text], DNN, RF, XGB, and SVM identified 9, 1, 2,  and 4 significant features, respectively. In a comparison of accuracy (Accuracy), the area under the curve (AUC), and the precision-recall area under the curve (PR-AUC), the permutation feature importance test for DNN model was the most powerful framework for identifying significant features and outperformed other methods, i.e., RF, XGB, and SVM, with Accuracy, AUC, and PR-AUC as 0.915, 0.794, and 0.974, respectively.

CONCLUSION

These results indicate that the PermFIT-DNN framework robustly identifies significant clinical features associated with TBI status and improves prediction performance. The findings could be used to inform the development of clinical decision tools designed to inform triage decisions.

Remote assessment of cognition and quality of life following radiotherapy for nasopharyngeal carcinoma: deep-learning-based predictive models and MRI correlates

PURPOSE

Irradiation of the brain regions from nasopharyngeal carcinoma (NPC) radiotherapy (RT) is frequently unavoidable, which may result in radiation-induced cognitive deficit. Using deep learning (DL), the study aims to develop prediction models in predicting compromised cognition in patients following NPC RT using remote assessments and determine their relation to the quality of life (QoL) and MRI changes.

METHODS

Seventy patients (20-76 aged) with MRI imaging (pre- and post-RT (6 months-1 year)) and complete cognitive assessments were recruited. Hippocampus, temporal lobes (TLs), and cerebellum were delineated and dosimetry parameters were extracted. Assessments were given post-RT via telephone (Telephone Interview Cognitive Status (TICS), Telephone Montreal Cognitive Assessment (T-MoCA), Telephone Mini Addenbrooke's Cognitive Examination (Tele-MACE), and QLQ-H&N 43). Regression and deep neural network (DNN) models were used to predict post-RT cognition using anatomical and treatment dose features.

RESULTS

Remote cognitive assessments were inter-correlated (r > 0.9). TLs showed significance in pre- and post-RT volume differences and cognitive deficits, that are correlated with RT-associated volume atrophy and dose distribution. Good classification accuracy based on DNN area under receiver operating curve (AUROC) for cognitive prediction (T-MoCA AUROC = 0.878, TICS AUROC = 0.89, Tele-MACE AUROC = 0.919).

CONCLUSION

DL-based prediction models assessed using remote assessments can assist in predicting cognitive deficit following NPC RT. Comparable results of remote assessments in assessing cognition suggest its possibility in replacing standard assessments.

IMPLICATIONS FOR CANCER SURVIVORS

Application of prediction models in individual patient enables tailored interventions to be provided in managing cognitive changes following NPC RT.

RegVar: Tissue-specific Prioritization of Non-coding Regulatory Variants

Non-coding genomic variants constitute the majority of trait-associated genome variations; however, the identification of functional non-coding variants is still a challenge in human genetics, and a method for systematically assessing the impact of regulatory variants on gene expression and linking these regulatory variants to potential target genes is still lacking. Here, we introduce a deep neural network (DNN)-based computational framework, RegVar, which can accurately predict the tissue-specific impact of non-coding regulatory variants on target genes. We show that by robustly learning the genomic characteristics of massive variant-gene expression associations in a variety of human tissues, RegVar vastly surpasses all current non-coding variant prioritization methods in predicting regulatory variants under different circumstances. The unique features of RegVar make it an excellent framework for assessing the regulatory impact of any variant on its putative target genes in a variety of tissues. RegVar is available as a web server at https://regvar.omic.tech/.

A deep generative adversarial network capturing complex spiral waves in disinhibited circuits of the cerebral cortex

BACKGROUND

In the cerebral cortex, disinhibited activity is characterized by propagating waves that spread across neural tissue. In this pathological state, a widely reported form of activity are spiral waves that travel in a circular pattern around a fixed spatial locus termed the center of mass. Spiral waves exhibit stereotypical activity and involve broad patterns of co-fluctuations, suggesting that they may be of lower complexity than healthy activity.

RESULTS

To evaluate this hypothesis, we performed dense multi-electrode recordings of cortical networks where disinhibition was induced by perfusing a pro-epileptiform solution containing 4-Aminopyridine as well as increased potassium and decreased magnesium. Spiral waves were identified based on a spatially delimited center of mass and a broad distribution of instantaneous phases across electrodes. Individual waves were decomposed into "snapshots" that captured instantaneous neural activation across the entire network. The complexity of these snapshots was examined using a measure termed the participation ratio. Contrary to our expectations, an eigenspectrum analysis of these snapshots revealed a broad distribution of eigenvalues and an increase in complexity compared to baseline networks. A deep generative adversarial network was trained to generate novel exemplars of snapshots that closely captured cortical spiral waves. These synthetic waves replicated key features of experimental data including a tight center of mass, a broad eigenvalue distribution, spatially-dependent correlations, and a high complexity. By adjusting the input to the model, new samples were generated that deviated in systematic ways from the experimental data, thus allowing the exploration of a broad range of states from healthy to pathologically disinhibited neural networks.

CONCLUSIONS

Together, results show that the complexity of population activity serves as a marker along a continuum from healthy to disinhibited brain states. The proposed generative adversarial network opens avenues for replicating the dynamics of cortical seizures and accelerating the design of optimal neurostimulation aimed at suppressing pathological brain activity.

Improvement of multi-task learning by data enrichment: application for drug discovery

Multi-task learning in deep neural networks has become a topic of growing importance in many research fields, including drug discovery. However, applying multi-task learning poses new challenges in improving prediction performance. This study investigated the potential of training data enrichment to enhance multi-task model prediction quality in drug discovery. The study evaluated four scenarios with varying degrees of information capacity of the training data and applied two types of test data to evaluate prediction performance. We used three datasets: ViralChEMBL, which consisted of binary activities of compounds against viral species, was applied for the classification task; pQSAR(159) and pQSAR(4267), which consisted of bio-activities of compounds and assays from the research of the profile-QSAR method, were applied for regression tasks. We built multi-task models based on the feed-forward DNNs using the PyTorch framework. Our findings showed that training data enrichment could be an effective means of enhancing prediction performance in multi-task learning, but the degree of improvement depends on the quality of the training data. The more unique compounds and targets the training data included, the more new compound-target interactions are required for prediction improvement. Also, we found out that even using multi-task learning, one could not predict the interactions of compounds that are highly dissimilar from those used for model training. The study provides some recommendations for effectively employing multi-task learning in drug discovery to improve prediction accuracy and facilitate the discovery of novel drug candidates.

Real-time deep neural network-based automatic bowel gas segmentation on X-ray images for particle beam treatment

Since particle beam distribution is vulnerable to change in bowel gas because of its low density, we developed a deep neural network (DNN) for bowel gas segmentation on X-ray images. We used 6688 image datasets from 209 cases as training data, 736 image datasets from 23 cases as validation data and 102 image datasets from 51 cases as test data (total 283 cases). For the training data, we prepared three types of digitally reconstructed radiographic (DRR) images (all-density, bone and gas) by projecting the treatment planning CT image data. However, the real X-ray images acquired in the treatment room showed low contrast that interfered with manual delineation of bowel gas. Therefore, we used synthetic X-ray images converted from DRR images in addition to real X-ray images.We evaluated DNN segmentation accuracy for the synthetic X-ray images using Intersection over Union, recall, precision, and the Dice coefficient, which measured 0.708 ± 0.208, 0.832 ± 0.170, 0.799 ± 0.191, and 0.807 ± 0.178, respectively. The evaluation metrics for the real X-images were less accurate than those for the synthetic X-ray images (0.408 ± 0237, 0.685 ± 0.326, 0.490 ± 0272, and 0.534 ± 0.271, respectively). Computation time was 29.7 ± 1.3 ms/image. Our DNN appears useful in increasing treatment accuracy in particle beam therapy.

Pollution level mapping of heavy metal in soil for ground-airborne hyperspectral data with support vector machine and deep neural network: A case study of Southwestern Xiong'an, China

Heavy metal in soil is a significant issue with the urban development in China, and traditional ground spectra are difficult to satisfy the demands for heavy metal monitoring and assessment in large-scale areas. In the paper, ground-airborne hyperspectral data is utilized to analyze the pollution level of heavy metal, 423 soil samples and corresponding ground spectra are collected synchronously with airborne hyperspectral image acquisition in Southwestern Xiong'an, China. Among them, support vector machine (SVM) is utilized to predict the concentration of independent samples, deep neural network (DNN) is aimed to estimate the spatial distribution of concentration with airborne image scenes. Finally, the pollution level is generated by the Softmax function, and it is defined by the risk control standard of heavy metals. The ground spectra and airborne image are closely integrated by the proposed method, the pollution situation is directly evaluated by ground-airborne hyperspectral data and indirectly evaluated by the concentration of local space, and the mapping results are believed to provide constructive advices about environmental protection.

Bio-medical imaging (X-ray, CT, ultrasound, ECG), genome sequences applications of deep neural network and machine learning in diagnosis, detection, classification, and segmentation of COVID-19: a Meta-analysis & systematic review

This review investigates how Deep Machine Learning (DML) has dealt with the Covid-19 epidemic and provides recommendations for future Covid-19 research. Despite the fact that vaccines for this epidemic have been developed, DL methods have proven to be a valuable asset in radiologists' arsenals for the automated assessment of Covid-19. This detailed review debates the techniques and applications developed for Covid-19 findings using DL systems. It also provides insights into notable datasets used to train neural networks, data partitioning, and various performance measurement metrics. The PRISMA taxonomy has been formed based on pretrained(45 systems) and hybrid/custom(17 systems) models with radiography modalities. A total of 62 systems with respect to X-ray(32), CT(19), ultrasound(7), ECG(2), and genome sequence(2) based modalities as taxonomy are selected from the studied articles. We originate by valuing the present phase of DL and conclude with significant limitations. The restrictions contain incomprehensibility, simplification measures, learning from incomplete labeled data, and data secrecy. Moreover, DML can be utilized to detect and classify Covid-19 from other COPD illnesses. The proposed literature review has found many DL-based systems to fight against Covid19. We expect this article will assist in speeding up the procedure of DL for Covid-19 researchers, including medical, radiology technicians, and data engineers.

Fake news detection: deep semantic representation with enhanced feature engineering

Due to the widespread use of social media, people are exposed to fake news and misinformation. Spreading fake news has adverse effects on both the general public and governments. This issue motivated researchers to utilize advanced natural language processing concepts to detect such misinformation in social media. Despite the recent research studies that only focused on semantic features extracted by deep contextualized text representation models, we aim to show that content-based feature engineering can enhance the semantic models in a complex task like fake news detection. These features can provide valuable information from different aspects of input texts and assist our neural classifier in detecting fake and real news more accurately than using semantic features. To substantiate the effectiveness of feature engineering besides semantic features, we proposed a deep neural architecture in which three parallel convolutional neural network (CNN) layers extract semantic features from contextual representation vectors. Then, semantic and content-based features are fed to a fully connected layer. We evaluated our model on an English dataset about the COVID-19 pandemic and a domain-independent Persian fake news dataset (TAJ). Our experiments on the English COVID-19 dataset show 4.16% and 4.02% improvement in accuracy and f1-score, respectively, compared to the baseline model, which does not benefit from the content-based features. We also achieved 2.01% and 0.69% improvement in accuracy and f1-score, respectively, compared to the state-of-the-art results reported by Shifath et al. (A transformer based approach for fighting covid-19 fake news, arXiv preprint arXiv:2101.12027, 2021). Our model outperformed the baseline on the TAJ dataset by improving accuracy and f1-score metrics by 1.89% and 1.74%, respectively. The model also shows 2.13% and 1.6% improvement in accuracy and f1-score, respectively, compared to the state-of-the-art model proposed by Samadi et al. (ACM Trans Asian Low-Resour Lang Inf Process, https://doi.org/10.1145/3472620, 2021).

A survey on recent trends in deep learning for nucleus segmentation from histopathology images

Nucleus segmentation is an imperative step in the qualitative study of imaging datasets, considered as an intricate task in histopathology image analysis. Segmenting a nucleus is an important part of diagnosing, staging, and grading cancer, but overlapping regions make it hard to separate and tell apart independent nuclei. Deep Learning is swiftly paving its way in the arena of nucleus segmentation, attracting quite a few researchers with its numerous published research articles indicating its efficacy in the field. This paper presents a systematic survey on nucleus segmentation using deep learning in the last five years (2017-2021), highlighting various segmentation models (U-Net, SCPP-Net, Sharp U-Net, and LiverNet) and exploring their similarities, strengths, datasets utilized, and unfolding research areas.

pH-dependent solubility prediction for optimized drug absorption and compound uptake by plants

Aqueous solubility is the most important physicochemical property for agrochemical and drug candidates and a prerequisite for uptake, distribution, transport, and finally the bioavailability in living species. We here present the first-ever direct machine learning models for pH-dependent solubility in water. For this, we combined almost 300000 data points from 11 solubility assays performed over 24 years and over one million data points from lipophilicity and melting point experiments. Data were split into three pH-classes - acidic, neutral and basic - , representing the conditions of stomach and intestinal tract for animals and humans, and phloem and xylem for plants. We find that multi-task neural networks using ECFP-6 fingerprints outperform baseline random forests and single-task neural networks on the individual tasks. Our final model with three solubility tasks using the pH-class combined data from different assays and five helper tasks results in root mean square errors of 0.56 log units overall (acidic 0.61; neutral 0.52; basic 0.54) and Spearman rank correlations of 0.83 (acidic 0.78; neutral 0.86; basic 0.86), making it a valuable tool for profiling of compounds in pharmaceutical and agrochemical research. The model allows for the prediction of compound pH profiles with mean and median RMSE per molecule of 0.62 and 0.56 log units.

Validation of an artificial intelligence platform for the guidance of safe laparoscopic cholecystectomy

BACKGROUND

Many surgical adverse events, such as bile duct injuries during laparoscopic cholecystectomy (LC), occur due to errors in visual perception and judgment. Artificial intelligence (AI) can potentially improve the quality and safety of surgery, such as through real-time intraoperative decision support. GoNoGoNet is a novel AI model capable of identifying safe ("Go") and dangerous ("No-Go") zones of dissection on surgical videos of LC. Yet, it is unknown how GoNoGoNet performs in comparison to expert surgeons. This study aims to evaluate the GoNoGoNet's ability to identify Go and No-Go zones compared to an external panel of expert surgeons.

METHODS

A panel of high-volume surgeons from the SAGES Safe Cholecystectomy Task Force was recruited to draw free-hand annotations on frames of prospectively collected videos of LC to identify the Go and No-Go zones. Expert consensus on the location of Go and No-Go zones was established using Visual Concordance Test pixel agreement. Identification of Go and No-Go zones by GoNoGoNet was compared to expert-derived consensus using mean F1 Dice Score, and pixel accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).

RESULTS

A total of 47 frames from 25 LC videos, procured from 3 countries and 9 surgeons, were annotated simultaneously by an expert panel of 6 surgeons and GoNoGoNet. Mean (± standard deviation) F1 Dice score were 0.58 (0.22) and 0.80 (0.12) for Go and No-Go zones, respectively. Mean (± standard deviation) accuracy, sensitivity, specificity, PPV and NPV for the Go zones were 0.92 (0.05), 0.52 (0.24), 0.97 (0.03), 0.70 (0.21), and 0.94 (0.04) respectively. For No-Go zones, these metrics were 0.92 (0.05), 0.80 (0.17), 0.95 (0.04), 0.84 (0.13) and 0.95 (0.05), respectively.

CONCLUSIONS

AI can be used to identify safe and dangerous zones of dissection within the surgical field, with high specificity/PPV for Go zones and high sensitivity/NPV for No-Go zones. Overall, model prediction was better for No-Go zones compared to Go zones. This technology may eventually be used to provide real-time guidance and minimize the risk of adverse events.

Mapping soil available copper content in the mine tailings pond with combined simulated annealing deep neural network and UAV hyperspectral images

Improper discharge of slag from mining will pollute the surrounding soil, thereby affecting the ecology and becoming an important global problem. The available copper (ACu) content in polluted soil is an important factor affecting plant growth and development. When investigating a large area of soil with ACu, manual sampling by points and inspection are mainly used, due to the heterogeneity of soil, the efficiency and accuracy are lower. The Unmanned aerial vehicle (UAV) equipped with a hyperspectral sensor as a remote sensing technology is widely used in soil indicator monitoring because of its rapid and convenience. Meanwhile, using the relationship between soil organic matter and available copper has the potential to predict available copper. In this study, we selected the study area with tailings area in the Jianghan Plain of China and used a UAV equipped with a hyperspectral sensor to predict ACu and soil organic matter (SOM) in the soil with two datasets. Firstly, 74 soil samples were collected in the study area, and the ACu and SOM of the soil samples were determined. Second, a hyperspectral image of the study area is obtained using a UAV equipped with a hyperspectral sensor. Thirdly, we combine hyperspectral data with competitive adaptive reweighted sampling (CARS) to obtain feature bands and utilize simulated annealing deep neural network (SA-DNN) to generate estimation models. Finally, maps of the distribution of ACu and SOM in the area were generated using the model. In two datasets, the model of ACu with R² values both are 0.89, and R² on the model of SOM is 0.89 and 0.88. The results show that the combination of UAV hyperspectral imagery with the SA-DNN model has good performance in the prediction of organic matter and available copper, which is helpful for soil environmental monitoring.

Tucker network: Expressive power and comparison

Deep neural networks have achieved great success in solving many machine learning and computer vision problems. In this paper, we propose a deep neural network called the Tucker network derived from the Tucker format and analyze its expressive power. The results demonstrate that the Tucker network has exponentially higher expressive power than the shallow network. In other words, a shallow network with an exponential width is required to realize the same score function as that computed by the Tucker network. Moreover, we discuss the expressive power between the hierarchical Tucker tensor network (HT network) and the proposed Tucker network. To generalize the Tucker network into a deep version, we combine the hierarchical Tucker format and Tucker format to propose a deep Tucker tensor decomposition. Its corresponding deep Tucker network is presented. Experiments are conducted on three datasets: MNIST, CIFAR-10 and CIFAR-100. The results experimentally validate the theoretical results and show that the Tucker network and deep Tucker network have better performance than the shallow network and HT network.

End-to-end emotional speech recognition using acoustic model adaptation based on knowledge distillation

The end-to-end approach provides better performance in speech recognition compared to the traditional hidden Markov model-deep neural network (HMM-DNN)-based approach, but still shows poor performance in abnormal speech, especially emotional speech. The optimal solution is to build an acoustic model suitable for emotional speech recognition using only emotional speech data for each emotion, but it is impossible because it is difficult to collect sufficient amount of emotional speech data for each emotion. In this study, we propose a method to improve the emotional speech recognition performance by using the knowledge distillation technique that was originally introduced to decrease computational intensity of deep learning-based approaches by reducing the number of model parameters. In addition to its use as model compression, we employ this technique for model adaptation to emotional speech. The proposed method builds a basic model (referred to as a teacher model) with a number of model parameters using an amount of normal speech data, and then constructs a target model (referred to as a student model) with fewer model parameters using a small amount of emotional speech data (i.e., adaptation data). Since the student model is built with emotional speech data, it is expected to reflect the emotional characteristics of each emotion well. In the emotional speech recognition experiment, the student model maintained recognition performance regardless of the number of model parameters, whereas the teacher model degraded performance significantly as the number of parameters decreased, showing performance degradation of about 10% in word error rate. This result demonstrates that the student model serves as an acoustic model suitable for emotional speech recognition even though it does not require much emotional speech data.

A vision-based real-time traffic flow monitoring system for road intersections

In this study, a vision based real-time traffic flow monitoring system has been developed to extract statistics passes through the intersections. A novel object tracking and data association algorithms have been developed using the bounding-box properties to estimate the vehicle trajectories. Then, rich traffic flow information such as directional and total counting, instantaneous and average speed of vehicles are calculated from the predicted trajectories. During the study, various parameters that affect the accuracy of vision based systems are examined such as camera locations and angles that may cause occlusion or illusion problems. In the last part, sample video streams are processed using both Kalman filter and new centroid-based algorithm for comparative study. The results show that the new algorithm performs 9.18% better than Kalman filter approach in general.

Deep learning-based method for automatic resolution of gas chromatography-mass spectrometry data from complex samples

Modern gas chromatography-mass spectrometry (GC-MS) is the workhorse for the high-throughput profiling of volatile compounds in complex samples. It can produce a considerable amount of two-dimensional data, and automatic methods are required to distill chemical information from raw GC-MS data efficiently. In this study, we proposed an Automatic Resolution method (AutoRes) based on pseudo-Siamese convolutional neural networks (pSCNN) to extract the meaningful features swamped by the noises, baseline drifts, retention time shifts, and overlapped peaks. Two pSCNN models were trained with 400,000 augmented spectral pairs, respectively. They can predict the selective region (pSCNN1) and elution region (pSCNN2) of compounds in an untargeted manner. The accuracies of the pSCNN1 model and the pSCNN2 model on their test sets are 99.9% and 92.6%, respectively. Then, the chromatographic profile of each component was automatically resolved by full rank resolution (FRR) based on the predicted regions by these models. The performance of AutoRes was evaluated on the simulated and plant essential oil datasets. Compared to AMDIS and MZmine, AutoRes resolves more reasonable mass spectra, chromatograms, and peak areas to identify and quantify compounds. The average match scores of AutoRes (925 and 936) outperformed AMDIS (909 and 925) and MZmine (888 and 916) when resolving mass spectra from overlapped peaks on the Set Ⅰ and Set Ⅱ of plant essential oil dataset and matching them against the NIST17 library. It extracted peak areas and mass spectra automatically from 10 GC-MS files of plant essential oils, and the entire process was completed in 8 min without any prior information or manual intervention. It is implemented in Python and is available as an open-source package at https://github.com/dyjfan/AutoRes.

Machine learning-based prediction of COVID-19 mortality using immunological and metabolic biomarkers

COVID-19 mortality prediction Background COVID-19 has become a major global public health problem, despite prevention and efforts. The daily number of COVID-19 cases rapidly increases, and the time and financial costs associated with testing procedure are burdensome. Method To overcome this, we aim to identify immunological and metabolic biomarkers to predict COVID-19 mortality using a machine learning model. We included inpatients from Hong Kong's public hospitals between January 1, and September 30, 2020, who were diagnosed with COVID-19 using RT-PCR. We developed three machine learning models to predict the mortality of COVID-19 patients based on data in their electronic medical records. We performed statistical analysis to compare the trained machine learning models which are Deep Neural Networks (DNN), Random Forest Classifier (RF) and Support Vector Machine (SVM) using data from a cohort of 5,059 patients (median age = 46 years; 49.3% male) who had tested positive for COVID-19 based on electronic health records and data from 532,427 patients as controls. Result We identified top 20 immunological and metabolic biomarkers that can accurately predict the risk of mortality from COVID-19 with ROC-AUC of 0.98 (95% CI 0.96-0.98). Of the three models used, our result demonstrate that the random forest (RF) model achieved the most accurate prediction of mortality among COVID-19 patients with age, glomerular filtration, albumin, urea, procalcitonin, c-reactive protein, oxygen, bicarbonate, carbon dioxide, ferritin, glucose, erythrocytes, creatinine, lymphocytes, PH of blood and leukocytes among the most important biomarkers identified. A cohort from Kwong Wah Hospital (131 patients) was used for model validation with ROC-AUC of 0.90 (95% CI 0.84-0.92). Conclusion We recommend physicians closely monitor hematological, coagulation, cardiac, hepatic, renal and inflammatory factors for potential progression to severe conditions among COVID-19 patients. To the best of our knowledge, no previous research has identified important immunological and metabolic biomarkers to the extent demonstrated in our study.

Supplementary Information

The online version contains supplementary material available at 10.1186/s44247-022-00001-0.

An approach based on multivariate distribution and Gaussian copulas to predict groundwater quality using DNN models in a data scarce environment

Machine Learning models have become a fruitful tool in water resources modelling. However, it requires a significant amount of datasets for training and validation, which poses challenges in the analysis of data scarce environments, particularly for poorly monitored basins. In such scenarios, using Virtual Sample Generation (VSG) method is valuable to overcome this challenge in developing ML models. The main aim of this manuscript is to introduce a novel VSG based on multivariate distribution and Gaussian Copula called MVD-VSG whereby appropriate virtual combinations of groundwater quality parameters can be generated to train Deep Neural Network (DNN) for predicting Entropy Weighted Water Quality Index (EWQI) of aquifers even with small datasets. The MVD-VSG is original and was validated for its initial application using sufficient observed datasets collected from two aquifers. The validation results showed that from only 20 original samples, the MVD-VSG provided enough accuracy to predict EWQI with an NSE of 0.87. However the companion publication of this Method paper is El Bilali et al. [1]. •Development of MVD-VSG to generate virtual combinations of groundwater parameters in data scarce environment.•Training deep neural network to predict groundwater quality.•Validation of the method with sufficient observed datasets and sensitivity analysis.

Automated warfarin dose prediction for Asian, American, and Caucasian populations using a deep neural network

Existing warfarin dose prediction algorithms based on pharmacogenetics and clinical parameters have not been used clinically due to the absence of external validation, lack of assessment for clinical utility, and high risk of bias. Moreover, given the high degree of heterogeneity across different datasets used to develop these algorithms, it is unsurprising that prediction errors remain high, and dosing accuracy is dependent on specific ethnic populations. To circumvent these challenges, deep neural models are increasingly used to improve the precision and accuracy of warfarin dose predictions. Hence, this study sought to develop a deep learning-based model using a well-established curated dataset of over 6000 patients from the International Warfarin Pharmacogenomics Consortium (IWPC). Clinically-relevant input data such as physical attributes, medical conditions, concomitant medications, genotype status of functional warfarin genetic polymorphisms, and therapeutic INR were entered followed by applying a unique and robust training and validation method. The deep model yielded a low average mean absolute error (MAE) of 7.6 mg/week and a relatively low mean percentage of error of 40.9% in Asians, 14.2 mg/week MAE and 36.9% in African Americans, and 12.7 mg/week MAE and 45.4% mean percentage of error in White Caucasians. This model also resulted in 36.4% of all patients with a predicted dose within 20% of the administered dose. Hence, our proposed deep model provides an alternative to predicting warfarin dose in the clinical setting upon validation in ethnically-similar datasets.