Impact Analysis of Evaluation Task Setting on a Public Dataset for Rolling Bearing Diagnostics Using Deep Learning

This dissertation explores the critical role of loss functions in enhancing the predictive performance of deep machine learning models. Loss functions are an integral element of all the ongoing advances we witness daily in this domain. I design custom loss functions and their impacts on various machine learning tasks, particularly in computer vision. In the first stage of my research, I aim to improve the prediction performance of deep learning models by providing them with more precise feedback associated with task requirements. This led me to create the concept of assistive loss functions. My first proposed loss function, inspired by the Active Shape Model (ASM), is ASM Loss, designed to improve the detection of facial landmarks and head pose accuracy in facial images. Additionally, influenced by Knowledge Distillation (KD), I introduce KD-Loss, a loss designed specifically for regression tasks, focusing on detecting facial landmarks. Proceeding to the second phase of my research, I focus on regression problems. Building upon my prior work, I introduce Adaptive Coordinate-based Regression (ACR)-Loss, a context-aware loss function with face alignment implication. I provide a mechanism to dynamically measure how hard the localization of each landmark point is, and accordingly, ACR-Loss provides unique feedback to the model. Furthermore, I introduce Intensity-aware Categorical (IC)-Loss for detecting sagittal cervical spine landmark points in medical images. It effectively deploys characteristics from classification problems to enhance regression optimization. Both ACR-Loss and IC-Loss comprise contextual information, leading to a significant enhancement in localization precision. In the third phase, I turned my attention toward classification problems within the domain of deep machine learning. I introduce Adaptive Correlation-based (Ad-corre) Loss, designed to enhance the discriminative capabilities of deep learning models, with a specific focus on improving classification accuracy in automatic facial expression recognition. Furthermore, I introduce Informative (Info) Loss, a specialized loss function developed for medical applications, specifically aimed at distinguishing between Mild Cognitive Impairment and Normal Cognition groups based on transcriptions of interviews conducted as part of the I-CONECT study. Overall, I have proposed custom loss functions to improve the predictive performance in both regression and classification problems. Furthermore, I have introduced assistive loss functions, designed to enhance the performance of the main loss functions. The collective outcomes of this extensive exploration highlight the critical role of loss functions, adapting and optimizing them for various domains in deep machine learning. This research suggests and proposes a variety of methodologies for designing adaptive and context-aware loss functions within deep machine learning.

Emotion recognition (ER) is crucial for understanding human behaviours, social interactions, and psychological well-being. Electroencephalography (EEG) has emerged as a promising tool for capturing the neural correlates of emotions. This work is a systematic review of articles in ER using EEG signals. A total of 120 articles from 1041 articles were selected based on PRISMA guidelines using defined inclusion and exclusion criteria, published between 2018 and 2024. This article aims to provide an in-depth understanding of the current landscape of ER from EEG signals utilizing deep learning (DL). This review offers valuable guidance for researchers and practitioners seeking more refined and reliable emotion classification systems. To explore the effectiveness of DL models in EEG-based ER, several potential DL models, such as convolutional neural network, long short-term memory (LSTM), gated recurrent unit (GRU), hybrid bidirectional LSTM (BiLSTM), bidirectional GRU, and advanced DL models such as convolutional recurrent neural network and EEG-Conformer models are applied to two popular datasets, SEED and GAMEEMO, respectively, to depict the full process of ER. Additionally, the performance of DL models is also compared with the performance of basic machine learning (ML) models such as SVM, k-nearest neighbors, logistic regression, and boosting algorithms such as AdaBoost, XGBoost and LightGBM. Through extensive experiments and performance evaluations, the performance of different models when applied to the datasets mentioned above is compared. The accuracy, precision, recall, and F1-scores are analysed to determine the most effective model for EEG-based ER. The findings of this study demonstrate that the performance of hybrid DL models is more efficacious than that of ML models. The best-performing model (BiLSTM) classified the emotions, with an accuracy of 90.54% when applied to the GAMEEMO dataset. This research contributes to the growing body of literature on ER and provides insights into the feasibility of using EEG signals to understand emotional states, and presents a structured roadmap for future exploration. The findings can aid in the development of more accurate and reliable ER systems, which can have wide-ranging applications in psychology, social sciences, and human-computer interactions.

Disease prediction based on modeling the correlations between compounded indicator factors is a widely used technique in high incidence chronic disease prevention diagnosis. Predictive models based on personal health information have been developed historically by using simple regression fitting over relatively few factors. Regression approaches have been favored in previous prediction modeling approaches because they are simplest and do not assume any non-linearity in the model for contributions of the chosen factors. In practice, many factors are correlated and have underlying non-linear relationships to the predicted outcome. Deep learning offers a means to construct a more complex modeling approach, along with automation and adaptation. The aim of this paper is to assess the ability of a deep learning model to predict the heart disease incidence using a common benchmark dataset (University of California, Irvine (UCI) dataset). The performance of deep learning model has been compared with four popular machine learning models (two linear and two nonlinear) in predicting the incidence of heart disease using data from 567 participants from two cohorts taken from UCI database. The deep learning model was able to achieve the best accuracy of 94% and an AUC score of 0.964 when compared to other models. The performance of deep learning and nonlinear machine learning models was significantly better compared to the linear machine learning models with increase in the dataset size.

This paper presents a supervised learning scheme that employs key-frame extraction to enhance the performance of pre-trained deep learning models for object detection in surveillance videos. Developing supervised deep learning models requires a significant amount of annotated video frames as training data, which demands substantial human effort for preparation. Key frames, which encompass frames containing false negative or false positive objects, can introduce diversity into the training data and contribute to model improvements. Our proposed approach focuses on detecting false negatives by leveraging the motion information within video frames that contain the detected object region. Key-frame extraction significantly reduces the human effort involved in video frame extraction. We employ interactive labeling to annotate false negative video frames with accurate bounding boxes and labels. These annotated frames are then integrated with the existing training data to create a comprehensive training dataset for subsequent training cycles. Repeating the training cycles gradually improves the object detection performance of deep learning models to monitor a new environment. Experiment results demonstrate that the proposed learning approach improves the performance of the object detection model in a new operating environment, increasing the mean average precision (mAP@0.5) from 54% to 98%. Manual annotation of key frames is reduced by 81% through the proposed key-frame extraction method.

Deep learning-based construction equipment operators’ mental fatigue classification using wearable EEG sensor data

Automated segmentation of dental restorations using deep learning: exploring data augmentation techniques.

The diagnostic performance of deep learning model that simultaneously detecting and quantifying nigrosome-1 abnormality by using susceptibility map-weighted imaging (SMwI) remains unexplored. This study aimed to develop and validate a deep learning-based automatic quantification for nigral hyperintensity and a classification algorithm for neurodegenerative parkinsonism. We retrospectively collected 450 participants (210 with idiopathic Parkinson disease [IPD] and 240 individuals in the control group) for training data between November 2022 and May 2023, and 237 participants (168 with IPD, 58 with essential tremor, and 11 with drug-induced parkinsonism) for validation data between July 2021 and January 2022. SMwI data were reconstructed from multiecho gradient echo. Diagnostic performance for diagnosing IPD was assessed by using deep learning-based automatic quantification (Heuron NI) and classification (Heuron IPD) models. Reference standard for IPD was based on N-3-fluoropropyl-2-β-carbomethoxy-3-β-(4-iodophenyl) nortropane PET finding. Additionally, the correlation between the Hoehn and Yahr (H&Y) stage and volume of nigral hyperintensity in patients with IPD was assessed. Quantification of nigral hyperintensity by using Heuron NI showed an area under the curve (AUC) of 0.915 (95% CI, 0.872-0.947) and 0.928 (95% CI, 0.887-0.957) on the left and right, respectively. Classification of nigral hyperintensity abnormality by using Heuron IPD showed area under the curve of 0.967 (95% CI, 0.936-0.986) and 0.976 (95% CI, 0.948-0.992) on the left and right, respectively. H&Y score ≥3 showed smaller nigral hyperintensity volume (1.43 ± 1.19 mm3) compared with H&Y score 1-2.5 (1.98 ± 1.63 mm3; P = .008). Our deep learning-based model proves rapid, accurate automatic quantification of nigral hyperintensity, facilitating IPD diagnosis, symptom severity prediction, and patient stratification for personalized therapy. Further study is warranted to validate the findings across various clinical settings.

Pathological is crucial to cancer diagnosis. Usually, Pathologists draw their conclusion based on observed cell and tissue structure on histology slides. Rapid development in machine learning, especially deep learning have established robust and accurate classifiers. They are being used to analyze histopathological slides and assist pathologists in diagnosis. Most machine learning systems rely heavily on annotated data sets to gain experiences and knowledge to correctly and accurately perform various tasks such as classification and segmentation. This work investigates different granularity of annotations in histopathological data set including image-wise, bounding box, ellipse-wise, and pixel-wise to verify the influence of annotation in pathological slide on deep learning models. We design corresponding experiments to test classification and segmentation performance of deep learning models based on annotations with different annotation granularity. In classification, state-of-the-art deep learning-based classifiers perform better when trained by pixel-wise annotation dataset. On average, precision, recall and F1-score improves by 7.87%, 8.83% and 7.85% respectively. Thus, it is suggested that finer granularity annotations are better utilized by deep learning algorithms in classification tasks. Similarly, semantic segmentation algorithms can achieve 8.33% better segmentation accuracy when trained by pixel-wise annotations. Our study shows not only that finer-grained annotation can improve the performance of deep learning models, but also help extracts more accurate phenotypic information from histopathological slides. Intelligence systems trained on granular annotations may help pathologists inspecting certain regions for better diagnosis. The compartmentalized prediction approach similar to this work may contribute to phenotype and genotype association studies.

To identify the feasibility of deep learning-based diagnostic models for detecting and assessing lower-extremity fatigue fracture severity on plain radiographs. This retrospective study enrolled 1151 X-ray images (tibiofibula/foot: 682/469) of fatigue fractures and 2842 X-ray images (tibiofibula/foot: 2000/842) without abnormal presentations from two clinical centers. After labeling the lesions, images in a center (tibiofibula/foot: 2539/1180) were allocated at 7:1:2 for model construction, and the remaining images from another center (tibiofibula/foot: 143/131) for external validation. A ResNet-50 and a triplet branch network were adopted to construct diagnostic models for detecting and grading. The performances of detection models were evaluated with sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), while grading models were evaluated with accuracy by confusion matrix. Visual estimations by radiologists were performed for comparisons with models. For the detection model on tibiofibula, a sensitivity of 95.4%/85.5%, a specificity of 80.1%/77.0%, and an AUC of 0.965/0.877 were achieved in the internal testing/external validation set. The detection model on foot reached a sensitivity of 96.4%/90.8%, a specificity of 76.0%/66.7%, and an AUC of 0.947/0.911. The detection models showed superior performance to the junior radiologist, comparable to the intermediate or senior radiologist. The overall accuracy of the diagnostic model was 78.5%/62.9% for tibiofibula and 74.7%/61.1% for foot in the internal testing/external validation set. The deep learning-based models could be applied to the radiological diagnosis of plain radiographs for assisting in the detection and grading of fatigue fractures on tibiofibula and foot. • Fatigue fractures on radiographs are relatively difficult to detect, and apt to be misdiagnosed. • Detection and grading models based on deep learning were constructed on a large cohort of radiographs with lower-extremity fatigue fractures. • The detection model with high sensitivity would help to reduce the misdiagnosis of lower-extremity fatigue fractures.

Deep neural networks are widely used in the field of image processing for micromachines, such as in 3D shape detection in microelectronic high-speed dispensing and object detection in microrobots. It is already known that hyperparameters and their interactions impact neural network model performance. Taking advantage of the mathematical correlations between hyperparameters and the corresponding deep learning model to adjust hyperparameters intelligently is the key to obtaining an optimal solution from a deep neural network model. Leveraging these correlations is also significant for unlocking the “black box” of deep learning by revealing the mechanism of its mathematical principle. However, there is no complete system for studying the combination of mathematical derivation and experimental verification methods to quantify the impacts of hyperparameters on the performances of deep learning models. Therefore, in this paper, the authors analyzed the mathematical relationships among four hyperparameters: the learning rate, batch size, dropout rate, and convolution kernel size. A generalized multiparameter mathematical correlation model was also established, which showed that the interaction between these hyperparameters played an important role in the neural network’s performance. Different experiments were verified by running convolutional neural network algorithms to validate the proposal on the MNIST dataset. Notably, this research can help establish a universal multiparameter mathematical correlation model to guide the deep learning parameter adjustment process.

With the rapid development of deep learning technology, the demand for computing resources is increasing, and the accelerated optimization of hardware on artificial intelligence (AI) chip has become one of the key ways to solve this challenge. This paper aims to explore the hardware acceleration optimization strategy of deep learning model on AI chip to improve the training and inference performance of the model. In this paper, the method and practice of optimizing deep learning model on AI chip are deeply analyzed by comprehensively considering the hardware characteristics such as parallel processing ability, energy-efficient computing, neural network accelerator, flexibility and programmability, high integration and heterogeneous computing structure. By designing and implementing an efficient convolution accelerator, the computational efficiency of the model is improved. The introduction of energy-efficient computing effectively reduces energy consumption, which provides feasibility for the practical application of mobile devices and embedded systems. At the same time, the optimization design of neural network accelerator becomes the core of hardware acceleration, and deep learning calculation such as convolution and matrix operation are accelerated through special hardware structure, which provides strong support for the real-time performance of the model. By analyzing the actual application cases of hardware accelerated optimization in different application scenarios, this paper highlights the key role of hardware accelerated optimization in improving the performance of deep learning model. Hardware accelerated optimization not only improves the computing efficiency, but also provides efficient and intelligent computing support for AI applications in different fields.

A wind speed forecasting technique, using deep learning architectures based on long short-term memory (LSTM) model and bidirectional long short-term memory (BiLSTM) model is presented in this work. The coastal belts of the Indian peninsula are vulnerable to natural disasters like storm surges and inundations due to cyclones each year. The wind speed is a major parameter for analyzing extreme weather events. Prediction using numerical models is not efficient enough due to the irregular patterns in the data and, thus, deep neural network models involving many layers have been tested. The shallow feed-forward model has also been considered along with deep learning models to estimate future values from past data. The present work employs a comparison study of different models to forecast wind speed time series at two locations in the Bay of Bengal and the Arabian Sea, respectively, having different dynamics and randomness. For training the models, daily wind speed data are considered for the period 2006–2017 and an independent validation set is chosen comprising 2018 wind speed data to check the accuracy. To evaluate forecast efficiency among different network models fitted to given time series, mean square error (MSE) and root mean square error (RMSE) have been computed. Multiple experiments are conducted with different hidden unit values and epoch values to obtain the minimum error. Regression equations generated may be used for forecasting future time series. The BiLSTM model connecting hidden states of opposite directions proved to be most efficient for the wind speed forecasting in different regions.

Area under the curve may hide poor generalisation to external datasets

Soil temperature (ST) plays an important role in agriculture and other fields, and has a close relationship with plant growth and development. Therefore, accurate ST prediction methods are widely needed. Deep learning (DL) models have been widely applied for ST prediction. However, the traditional DL models may fail to capture the spatiotemporal relationship due to its complex dependency under different related hydrologic variables. Hence, the DL models with Ensemble Empirical Mode Decomposition (EEMD) are proposed in this study. The proposed models can capture more complex spatiotemporal relationship after decomposing the ST into different intrinsic mode functions. Therefore, the performance of models is further improved. The results show that the performance of DL models with EEMD are better than that of corresponding DL models without EEMD. Moreover, EEMD-Conv3d has the best performance among all the experimental models. It has the highest R2 ranging from 0.9826 to 0.9893, the lowest RMSE ranging from 1.3096 to 1.6497 and the lowest MAE ranging from 0.9656 to 1.2056 in predicting ST at the lead time from one to five days. In addition, the lines between predicted ST and observed ST are closer to the ideal line (y = x) than other DL models. The results show that our EEMD-Conv3D can better capture spatiotemporal correlation and is an applicable method for predicting spatiotemporal ST.

Aims: This study aimed to explore the effect of training set diversity on the performance of deep learning models for predicting early gastric cancer (EGC) under endoscopy. Methods: Images of EGC and non-cancerous lesions under narrow-band imaging (ME-NBI) and magnifying blue laser imaging (ME-BLI) were retrospectively collected. Training set 1 was composed of 150 non-cancerous and 309 EGC ME-NBI images, training set 2 was composed of 1505 non-cancerous and 309 EGC ME-BLI images, and training set 3 was the combination of training set 1 and 2. Test set 1 was composed of 376 non-cancerous and 1052 EGC ME-NBI images, test set 2 consisted of 529 non-cancerous and 71 EGC ME-BLI images, and test set 3 was the combination of test set 1 and test set 2. Three deep learning models, convolutional neural network (CNN) 1, CNN 2 and CNN 3 (CNN 1, CNN 2 and CNN 3 were independently trained using training set 1, training set 2 and training set 3, respectively), were constructed, and their performances on each test set were respectively evaluated. One hundred and thirty-eight ME-NBI videos and 17 ME-BLI videos were further collected to evaluate and compare the performance of each model in real time. Results: On the whole, the performance of CNN 3 was the best. The accuracy (Acc), sensitivity (Sn), specificity (Sp) and area under the curve (AUC) of test set 1 in CNN 3 were 87.89% (1255/1428), 90.96% (342/376), 86.79% (913/1052) and 94.60%, respectively. The Acc, Sn, Sp and AUC of test set 2 in CNN 3 were 95% (570/600), 97.92% (518/529), 73.24% (52/71) and 90.93% respectively. The Acc, Sn, Sp and AUC of test set 3 in CNN 3 were 89.99% (1825/2028), 95.03% (860/905), 85.93% (965/1123) and 94.89%, respectively. The performance of CNN 3 was also the best in videos test set. The Acc, Sn and Sp of videos test set in CNN 3 were 91.03% (142/156), 90.58% (125/138) and 94.44% (17/18), respectively. Conclusions: The deep learning model with the most diverse training data has the best diagnostic effect.