Abstract Background: The diagnosis of most cancers is made by a board-certified pathologist based on a tissue biopsy under the microscope. Recent research reveals a high discordance between individual pathologists. For melanoma, the literature reports on 25e26% of discordance for classifying a benign nevus versus malignant melanoma. A recent study indicated the potential of deep learning to lower these discordances. However, the performance of deep learning in classifying histopathologic melanoma images was never compared directly to human experts. The aim of this study is to perform such a first direct comparison. Methods: A total of 695 lesions were classified by an expert histopathologist in accordance with current guidelines (350 nevi/345 melanoma). Only the haematoxylin & eosin (H&E) slides of these lesions were digitalised via a slide scanner and then randomly cropped. A total of 595 of the resulting images were used to train a convolutional neural network (CNN). The additional 100 H&E image sections were used to test the results of the CNN in comparison to 11 histopathologists. Three combined McNemar tests comparing the results of the CNNs test runs in terms of sensitivity, specificity and accuracy were predefined to test for significance (p < 0.05). Findings: The CNN achieved a mean sensitivity/specificity/accuracy of 76%/60%/68% over 11 test runs. In comparison, the 11 pathologists achieved a mean sensitivity/specificity/accuracy of 51.8%/66.5%/59.2%. Thus, the CNN was significantly (p Z 0.016) superior in classifying the cropped images. Interpretation: With limited image information available, a CNN was able to outperform 11 histopathologists in the classification of histopathological melanoma images and thus shows promise to assist human melanoma diagnoses.
KEYWORDS : Melanoma | Pathology | Histopathology | Deep learning | Artificial intelligence

Most common types of defects for composite are debond and delamination. It is difficult to detect the inner defects on a complex shaped specimen by using conventional optical thermography nondestructive testing (NDT) methods. In this paper, a hybrid of spatial and temporal deep learning architecture for automatic thermography defects detection is proposed. The integration of cross network learning strategy has the capability to significantly minimize the uneven illumination and enhance the detection rate. The probability of detection (POD) has been derived to measure the detection results and this is coupled with comparison studies to verify the efficacy of the proposed method. The results show that visual geometry group-Unet (VGG-Unet) cross learning structure can significantly improve the contrast between the defective and non-defective regions. In addition, investigation of different feature extraction methods in which embedded in deep learning is conducted to optimize the learning structure. To investigate the efficacy and robustness of the proposed method, experimental studies have been carried out for inner debond defects on both regular and irregular shaped carbon fiber reinforced polymer (CFRP) specimens.
Keywords: Deep learning | Segmentation | Thermography defect detection | Nondestructive testing

Machine learning (ML) plays the core function to intellectualize the transportation systems. Recent years have witnessed the advent and prevalence of deep learning which has provoked a storm in ITS (Intelligent Transportation Systems). Consequently, traditional ML models in many applications have been replaced by the new learning techniques and the landscape of ITS is being reshaped. Under such perspective, we provide a comprehensive survey that focuses on the utilization of deep learning models to enhance the intelligence level of transportation systems. By organizing multiple dozens of relevant works that were originally scattered here and there, this survey attempts to provide a clear picture of how various deep learning models have been applied in multiple transportation applications.
Keywords: Deep learning | Transportation systems | Survey

The current paper proposes a novel stochastic procedure for modelling car-following behaviours on a multi-lane motorway. We develop an integrated multi-lane stochastic continuous car-following model where a deep learning architecture is used to estimate a probability of lanechanging (LC) manoeuvres. To the best of our knowledge, this work is among the very few papers which exploit deep learning to model driving behaviour on a multi-lane road. The objective of this study is to establish a coupled stochastic continuous multi-lane car-following model using Langevin equations to cope with probabilistic characteristics of LC manoeuvres. In particular, a stochastic volatility, derived from LC manoeuvres is introduced in a multi-lane stochastic optimal velocity model (SOVM). In additions, Convolutional Neural Network (CNN) is applied to estimate a probability of LC manoeuvres in the integrated multi-lane car-following model. Furthermore, imaged second-based trajectories of the lane-changer and surrounding vehicles are used to identify whether LC manoeuvres occur by using the CNN. Finally, the proposed method is validated using a real-world high-resolution vehicle trajectory dataset. The results indicate that the prediction of the integrated SOVM is almost identical to the observed trajectories of the lanechangers and the following vehicles in the initial and the target lane. It has been found that the proposed multi-lane SOVM can tackle the unpredictable fluctuations in the velocity of the vehicles in the acceleration/deceleration zone.
Keywords: Stochastic car-following model | Deep learning | Lane-changing behaviour

In aerospace industry, gears are the most common parts of a mechanical transmission system. Gear pitting faults could cause the transmission system to crash and give rise to safety disaster. It is always a challenging problem to diagnose the gear pitting condition directly through the raw signal of vibration. In this paper, a novel method named augmented deep sparse autoencoder (ADSAE) is proposed. The method can be used to diagnose the gear pitting fault with relatively few raw vibration signal data. This method is mainly based on the theory of pitting fault diagnosis and creatively combines with both data augmentation ideology and the deep sparse autoencoder algorithm for the fault diagnosis of gear wear. The effectiveness of the proposed method is validated by experiments of six types of gear pitting conditions. The results show that the ADSAE method can effectively increase the network generalization ability and robustness with very high accuracy. This method can effectively diagnose different gear pitting conditions and show the obvious trend according to the severity of gear wear faults. The results obtained by the ADSAE method proposed in this paper are compared with those obtained by other common deep learning methods. This paper provides an important insight into the field of gear fault diagnosis based on deep learning and has a potential practical application value.
KEYWORDS : Deep learning | Gear pitting diagnosis | Gear teeth | Raw vibration signal | Semi-supervised learning | Sparse autoencoder