From Structure to Function: Biological Activity Prediction of Phytochemicals Using Molecular Fingerprints with Convolutional Neural Networks

Neural network-based approaches for biomedical relation classification: A review.

Pneumonia is a lung infection and causes the inflammation of the small air sacs (Alveoli) in one or both lungs. Proper and faster diagnosis of pneumonia at an early stage is imperative for optimal patient care. Currently, chest X-ray is considered as the best imaging modality for diagnosing pneumonia. However, the interpretation of chest X-ray images is challenging. To this end, we aimed to use an automated convolutional neural network-based transfer-learning approach to detect pneumonia in paediatric chest radiographs. Herein, an automated convolutional neural network-based transfer-learning approach using four different pre-trained models (i.e. VGG19, DenseNet121, Xception, and ResNet50) was applied to detect pneumonia in children (1-5 years) chest X-ray images. The performance of different proposed models for testing data set was evaluated using five performances metrics, including accuracy, sensitivity/recall, Precision, area under curve, and F1 score. All proposed models provide accuracy greater than 83.0% for binary classification. The pre-trained DenseNet121 model provides the highest classification performance of automated pneumonia classification with 86.8% accuracy, followed by Xception model with an accuracy of 86.0%. The sensitivity of the proposed models was greater than 91.0%. The Xception and DenseNet121 models achieve the highest classification performance with F1-score greater than 89.0%. The plotted area under curve of receiver operating characteristics of VGG19, Xception, ResNet50, and DenseNet121 models are 0.78, 0.81, 0.81, and 0.86, respectively. Our data showed that the proposed models achieve a high accuracy for binary classification. Transfer learning was used to accelerate training of the proposed models and resolve the problem associated with insufficient data. We hope that these proposed models can help radiologists for a quick diagnosis of pneumonia at radiology departments. Moreover, our proposed models may be useful to detect other chest-related diseases such as novel Coronavirus 2019. Herein, we used transfer learning as a machine learning approach to accelerate training of the proposed models and resolve the problem associated with insufficient data. Our proposed models achieved accuracy greater than 83.0% for binary classification.

Micro-pin fin heat sinks are widely used to cool miniature devices. The flow characteristics and cooling performance of these heat sinks are highly dependent on their geometric configuration. Previous studies have focused on optimizing the design so that the pressure drop decreases, while the heat transfer performance is maintained. However, limited numbers of geometries have been explored, mainly considering only homogeneous pin fin arrays. In this study, we propose a neural network-based regression approach called the flow-learned building block (FLBB) and develop an effective parametric study and optimization for micro-pin fin heat sinks including heterogeneous geometries. The prediction capabilities of the FLBB are verified by comparing the predicted results with direct numerical simulation results for various pitch distances, pin sizes, and arrangements at Reynolds numbers from 1 to 100. Furthermore, we demonstrate the applicability of the FLBB to different working fluids, quantified by the Prandtl number (0.71 ≤ Pr ≤ 5.86). Leveraging the reliable and effective prediction capabilities of our neural network-based approach, we perform parametric studies of micro-pin fin heat sinks for working fluids of air and water with the aim of minimizing the pump power and achieving uniform heat transfer along the pin fins.

Personal identification based on ECG signals has been a significant challenge. The performance of an ECG authentication system depends significantly on the features extracted and the classifier subsequently applied. Although recently the deep neural networks based approaches featuring adaptive feature extractions and inherent classifications have attracted attention, they usually require a substantial set of training data. Aiming at tackling these issues, this paper presents a convolutional neural network-based transfer learning approach. It includes transferring the big data-trained GoogLeNet model into our identification task, fine-tuning the model using the ‘finetune’ idea, and adding three adaptive layers behind the original feature layer. The proposed approach not only requires a small set of training data, but also obtains great performance.

To develop a proof-of-concept convolutional neural network (CNN) to synthesize T2 maps in right lateral femoral condyle articular cartilage from anatomic MR images by using a conditional generative adversarial network (cGAN). In this retrospective study, anatomic images (from turbo spin-echo and double-echo in steady-state scans) of the right knee of 4621 patients included in the 2004-2006 Osteoarthritis Initiative were used as input to a cGAN-based CNN, and a predicted CNN T2 was generated as output. These patients included men and women of all ethnicities, aged 45-79 years, with or at high risk for knee osteoarthritis incidence or progression who were recruited at four separate centers in the United States. These data were split into 3703 (80%) for training, 462 (10%) for validation, and 456 (10%) for testing. Linear regression analysis was performed between the multiecho spin-echo (MESE) and CNN T2 in the test dataset. A more detailed analysis was performed in 30 randomly selected patients by means of evaluation by two musculoskeletal radiologists and quantification of cartilage subregions. Radiologist assessments were compared by using two-sided t tests. The readers were moderately accurate in distinguishing CNN T2 from MESE T2, with one reader having random-chance categorization. CNN T2 values were correlated to the MESE values in the subregions of 30 patients and in the bulk analysis of all patients, with best-fit line slopes between 0.55 and 0.83. With use of a neural network-based cGAN approach, it is feasible to synthesize T2 maps in femoral cartilage from anatomic MRI sequences, giving good agreement with MESE scans.See also commentary by Yi and Fritz in this issue.Keywords: Cartilage Imaging, Knee, Experimental Investigations, Quantification, Vision, Application Domain, Convolutional Neural Network (CNN), Deep Learning Algorithms, Machine Learning Algorithms© RSNA, 2021.

The rampant propagation of rumors in online social networks leads to potential damage to society. This phenomenon has attracted significant attention to the researches on faster rumor detection. Most recent works on rumor detection adapted the neural network-based deep learning approaches because of its high success rate. Though neural network-based models have high success rates, these models are less efficient in finding rumors at the earliest. Such models require huge training data for better and accurate results. Unfortunately, the data available in the early stages of rumor propagation are sparse in nature. This nature makes rumor detection using neural networks a complex solution, therefore rendering less efficiency for early stage rumors. To address this issue, we have proposed a certainty-factor-based convolutional neural network (CNN) approach to efficiently classify events as rumor or not by leveraging the inherent features of the set of information in spite of data sparsity. The certainty-factor-based activation function requires a minimum number of training data to obtain a generalization. In the proposed approach, two parallel CNNs are employed for the rumor event classification task that efficiently utilizes the inherent features of information, such as temporal, content, and propagation features. A decision tree combines the outputs of both CNNs and provides the classification output. This rumor event classification approach is then compared with recent and well-known state-of-the-art rumor classification approaches, and the results prove that the proposed approach detects rumor efficiently and rapidly with minimal inputs compared to other approaches.

Physical rehabilitation is essential for a large number of patients around the world to recover from their disabilities. However, conventional methods of rehabilitation were expensive and difficult for majority of patients to access. This paper presents studies with neural network-based approaches that solve this problem are reviewed in this paper. The methods are divided into three main categories: stroke rehabilitation, injury rehabilitation and other rehabilitations. The common methods reviewed are Convolutional Neural Network (CNN), Support Vector Machine (SVM) and Recurrent Neural Network (RNN). There are various rehabilitation datasets reviewed in the paper, which all included pictures and videos of both patients and healthy people preforming a series of movements. The sensors used in experiments to capture patients movements are also summarized in this paper. The paper reviewed few algorithms able to model a 3D human skeleton based on the data collected by sensors. Evaluation metrics reviewed includes Discrete Movement Score, rule-based and template-based scoring methods. K-Nearest Neighbor (KNN) and Dynamic Time Warping (DTW) distance function were commonly used in template-based evaluation methods. The results in researches reviewed indicate that neural network-based rehabilitation methods are able to satisfy most demands, and improved efficiency, making it affordable and accessible to more patients.

The perceived texture directionality is an important, not fully explored image characteristic. In many applications texture directionality detection is of fundamental importance. Several approaches have been proposed, such as the fast Fourier-based method. We recently proposed a method based on the interpolated grey-level co-occurrence matrix (iGLCM), robust to image blur and noise but slower than the Fourier-based method. Here we test the applicability of convolutional neural networks (CNNs) to texture directionality detection. To obtain the large amount of training data required, we built a training dataset consisting of synthetic textures with known directionality and varying perturbation levels. Subsequently, we defined and tested shallow and deep CNN architectures. We present the test results focusing on the CNN architectures and their robustness with respect to image perturbations. We identify the best performing CNN architecture, and compare it with the iGLCM, the Fourier and the local gradient orientation methods. We find that the accuracy of CNN is lower, yet comparable to the iGLCM, and it outperforms the other two methods. As expected, the CNN method shows the highest computing speed. Finally, we demonstrate the best performing CNN on real-life images. Visual analysis suggests that the learned patterns generalize to real-life image data. Hence, CNNs represent a promising approach for texture directionality detection, warranting further investigation.

Interferometry Synthetic Aperture Radar (InSAR) is an advanced remote sensing technique for studying the earth’s surface topography and deformations; it is used to generate high-quality Digital Elevation Models (DEMs). DEMs are a crucial and primary input to various topographical quantification and modelling applications. The quality of input DEMs can be further improved using fusion methods, which combine multi-sensor or multi-temporal datasets intelligently to retrieve the best information from the input data. This research study is based on developing a Neural Network-based fusion approach for improving InSAR-based DEMs in plain and hilly terrain parts of India. The study areas comprise relatively plain terrain from Ghaziabad and hilly terrain of Dehradun and their surrounding regions. The training dataset consists of DEM elevations and derived topographic attributes like slope, aspect, topographic position index (TPI), terrain ruggedness index (TRI), and vector roughness measure (VRM) in different land use land cover classes of the study areas. The spaceborne altimetry ICESat-2 ATL08 photon data are used as a reference elevation. A Feed Forward Neural Network with a backpropagation algorithm is trained based on the prepared training samples. The trained model produces fused DEMs by learning the relationship between the input and target samples; this is used to predict elevations for the test areas. The accuracy of results from the models is assessed with TanDEM-X 90 m DEM. The fused DEMs show significant improvement in terms of RMSE (Root Mean Square Error) over the input DEMs with an improvement factor of 94.65% in plain areas and 82.62% in hilly areas. The study concludes that the ANN with its universal approximation property can significantly improve the fused DEM.

An neural network-based modelling approach is employed to predict the grain growth behaviour during continuous reheating. Using a significant data set containing critical information on the grain growth, a neural network based model has been trained. A compact set of process variables has been selected as the model inputs, based on expert knowledge as well as data analysis techniques. Ensemble modelling techniques have been used to improve model performance as well as to provide error bounds for prediction confidence. The resulting neural network model gives an impressive prediction performance, with the prediction error very close to the maximal measurement standard deviation. The neural network model has been tested on new grain growth data with more divergence in the reheating patterns, and gives a satisfactory prediction on these data as well. It is concluded that the developed grain growth model is capable of providing the initial microstructures for an integrated thermomechanical model, with a very fast computing speed.

A convolutional neural network-based approach to composite power system reliability evaluation

Fault diagnosis in photovoltaic (PV) arrays is essential in enhancing power output as well as the useful life span of a PV system. Severe faults such as Partial Shading (PS) and high impedance faults, low location mismatch, and the presence of Maximum Power Point Tracking (MPPT) make fault detection challenging in harsh environmental conditions. In this regard, there have been several attempts made by various researchers to identify PV array faults. However, most of the previous work has focused on fault detection and classification in only a few faulty scenarios. This paper presents a novel approach that utilizes deep two-dimensional (2-D) Convolutional Neural Networks (CNN) to extract features from 2-D scalograms generated from PV system data in order to effectively detect and classify PV system faults. An in-depth quantitative evaluation of the proposed approach is presented and compared with previous classification methods for PV array faults – both classical machine learning based and deep learning based. Unlike contemporary work, five different faulty cases (including faults in PS – on which no work has been done before in the machine learning domain) have been considered in our study, along with the incorporation of MPPT. We generate a consistent dataset over which to compare ours and previous approaches, to make for the first (to the best of our knowledge) comprehensive and meaningful comparative evaluation of fault diagnosis. It is observed that the proposed method involving fine-tuned pre-trained CNN outperforms existing techniques, achieving a high fault detection accuracy of 73.53%. Our study also highlights the importance of representative and discriminative features to classify faults (as opposed to the use of raw data), especially in the noisy scenario, where our method achieves the best performance of 70.45%. We believe that our work will serve to guide future research in PV system fault diagnosis.

Nowadays various applications are available for plant disease identification using images. Early-stage disease identification can reduce losses and cost in cultivation. Efficient image segmentation is required to improve the performance of plant disease identification. Selecting appropriate segmentation techniques to extract an accurate object of interest while preserving original image properties is a challenging task. This paper presents the principle of image segmentation covering different techniques from traditional thresholding to the latest convolutional neural network-based approach. We reviewed the selected paper based on image segmentation and crop disease identification. Various algorithms are grouped based on the working principles like edge-based, region-based, and combining both properties. Performance evaluation of these algorithms was carried out using factors like time required, accuracy, and similarity to the original image. Holistic image segmentation based on convolutional neural network, K means clustering, etc. algorithms applied to real field crop images. The grab cut algorithm proves very useful for real field crop image segmentation as it preserves original image properties. Combining region and boundary-based techniques and automating segmentation need to be explored in future research work.KeywordsMachine visionImage processingDeep learningK means clusteringGrabcut

Predicting accurate protein-ligand binding affinities is an important task in drug discovery but remains a challenge even with computationally expensive biophysics-based energy scoring methods and state-of-the-art deep learning approaches. Despite the recent advances in the application of deep convolutional and graph neural network-based approaches, it remains unclear what the relative advantages of each approach are and how they compare with physics-based methodologies that have found more mainstream success in virtual screening pipelines. We present fusion models that combine features and inference from complementary representations to improve binding affinity prediction. This, to our knowledge, is the first comprehensive study that uses a common series of evaluations to directly compare the performance of three-dimensional (3D)-convolutional neural networks (3D-CNNs), spatial graph neural networks (SG-CNNs), and their fusion. We use temporal and structure-based splits to assess performance on novel protein targets. To test the practical applicability of our models, we examine their performance in cases that assume that the crystal structure is not available. In these cases, binding free energies are predicted using docking pose coordinates as the inputs to each model. In addition, we compare these deep learning approaches to predictions based on docking scores and molecular mechanic/generalized Born surface area (MM/GBSA) calculations. Our results show that the fusion models make more accurate predictions than their constituent neural network models as well as docking scoring and MM/GBSA rescoring, with the benefit of greater computational efficiency than the MM/GBSA method. Finally, we provide the code to reproduce our results and the parameter files of the trained models used in this work. The software is available as open source at https://github.com/llnl/fast. Model parameter files are available at ftp://gdo-bioinformatics.ucllnl.org/fast/pdbbind2016_model_checkpoints/.

In this paper, a mosque recognition system based on the Convolutional Neural Networks (CNN) and the Laplacian Pyramid is proposed. CNN networks have the ability to autonomously obtain and extract from images the features that can assist in the decision making. However, due to some factors such as different environmental conditions as well as different capturing conditions, the learned features by the CNN network can be misleading if trained using an incomplete dataset. Thus, in this paper, the preferred features which are invariant to different capturing conditions are first extracted from raw images using the Laplacian pyramid. This is done by extracting the lowest level of the Laplacian pyramid of each image. The first Laplacian level is a high-frequency representation of an image that emphasizes the edges of an image, which are invariant features to various lighting conditions. Next, these high-frequency images are fed to the CNN network instead of the raw images. Experimental results have proved the efficiency of the proposed technique, where the accuracy of the proposed scheme has achieved an accuracy percentage of 97.25% surpassing other conventional machine learning techniques such as Muti-Layer Perceptron (MLP) and Support Vector Machine (SVM).