A deep learning fusion network trained with clinical and high-frequency ultrasound images in the multi-classification of skin diseases in comparison with dermatologists: a prospective and multicenter study.

Background: Clinical appearance and high-frequency ultrasound (HFUS) are indispensable for diagnosing skin diseases by providing internal and external information. However, their complex combination brings challenges for primary care physicians and dermatologists. Thus, we developed a deep multimodal fusion network (DMFN) model combining analysis of clinical close-up and HFUS images for binary and multiclass classification in skin diseases. Methods: Between Jan 10, 2017, and Dec 31, 2020, the DMFN model was trained and validated using 1269 close-ups and 11,852 HFUS images from 1351 skin lesions. The monomodal convolutional neural network (CNN) model was trained and validated with the same close-up images for comparison. Subsequently, we did a prospective and multicenter study in China. Both CNN models were tested prospectively on 422 cases from 4 hospitals and compared with the results from human raters (general practitioners, general dermatologists, and dermatologists specialized in HFUS). The performance of binary classification (benign vs. malignant) and multiclass classification (the specific diagnoses of 17 types of skin diseases) measured by the area under the receiver operating characteristic curve (AUC) were evaluated. This study is registered with www.chictr.org.cn (ChiCTR2300074765). Findings: The performance of the DMFN model (AUC, 0.876) was superior to that of the monomodal CNN model (AUC, 0.697) in the binary classification (P = 0.0063), which was also better than that of the general practitioner (AUC, 0.651, P = 0.0025) and general dermatologists (AUC, 0.838; P = 0.0038). By integrating close-up and HFUS images, the DMFN model attained an almost identical performance in comparison to dermatologists (AUC, 0.876 vs. AUC, 0.891; P = 0.0080). For the multiclass classification, the DMFN model (AUC, 0.707) exhibited superior prediction performance compared with general dermatologists (AUC, 0.514; P = 0.0043) and dermatologists specialized in HFUS (AUC, 0.640; P = 0.0083), respectively. Compared to dermatologists specialized in HFUS, the DMFN model showed better or comparable performance in diagnosing 9 of the 17 skin diseases. Interpretation: The DMFN model combining analysis of clinical close-up and HFUS images exhibited satisfactory performance in the binary and multiclass classification compared with the dermatologists. It may be a valuable tool for general dermatologists and primary care providers. Funding: This work was supported in part by the National Natural Science Foundation of China and the Clinical research project of Shanghai Skin Disease Hospital.

Unsupervised band selection of medical hyperspectral images guided by data gravitation and weak correlation.

BACKGROUND AND OBJECTIVE: Medical hyperspectral images (MHSIs) are used for a contact-free examination of patients without harmful radiation. However, high-dimensionality images contain large amounts of data that are sparsely distributed in a high-dimensional space, which leads to the "curse of dimensionality" (called Hughes' phenomenon) and increases the complexity and cost of data processing and storage. Hence, there is a need for spectral dimensionality reduction before the clinical application of MHSIs. Some dimensionality-reducing strategies have been proposed; however, they distort the data within MHSIs. METHODS: To compress dimensionality without destroying the original data structure, we propose a method that involves data gravitation and weak correlation-based ranking (DGWCR) for removing bands of noise from MHSIs while clustering signal-containing bands. Band clustering is done by using the connection centre evolution (CCE) algorithm and selecting the most representative bands in each cluster based on the composite force. The bands within the clusters are ranked using the new entropy-containing matrix, and a global ranking of bands is obtained by applying an S-shaped strategy. The source code is available at https://www.github.com/zhangchenglong1116/DGWCR. RESULTS: Upon feeding the reduced-dimensional images into various classifiers, the experimental results demonstrated that the small number of bands selected by the proposed DGWCR consistently achieved higher classification accuracy than the original data. Unlike other reference methods (e.g. the latest deep-learning-based strategies), DGWCR chooses the spectral bands with the least redundancy and greatest discrimination. CONCLUSION: In this study, we present a method for efficient band selection for MHSIs that alleviates the "curse of dimensionality". Experiments were validated with three MHSIs in the human brain, and they outperformed several other band selection methods, demonstrating the clinical potential of DGWCR.

A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images.

In recent years, the incidence of thyroid cancer has been increasing. Thyroid nodule detection is critical for both the detection and treatment of thyroid cancer. Convolutional neural networks (CNNs) have achieved good results in thyroid ultrasound image analysis tasks. However, due to the limited valid receptive field of convolutional layers, CNNs fail to capture long-range contextual dependencies, which are important for identifying thyroid nodules in ultrasound images. Transformer networks are effective in capturing long-range contextual information. Inspired by this, we propose a novel thyroid nodule detection method that combines the Swin Transformer backbone and Faster R-CNN. Specifically, an ultrasound image is first projected into a 1D sequence of embeddings, which are then fed into a hierarchical Swin Transformer. The Swin Transformer backbone extracts features at five different scales by utilizing shifted windows for the computation of self-attention. Subsequently, a feature pyramid network (FPN) is used to fuse the features from different scales. Finally, a detection head is used to predict bounding boxes and the corresponding confidence scores. Data collected from 2,680 patients were used to conduct the experiments, and the results showed that this method achieved the best mAP score of 44.8%, outperforming CNN-based baselines. In addition, we gained better sensitivity (90.5%) than the competitors. This indicates that context modeling in this model is effective for thyroid nodule detection.

Aerial scene understanding in the wild: Multi-scene recognition via prototype-based memory networks.

Aerial scene recognition is a fundamental visual task and has attracted an increasing research interest in the last few years. Most of current researches mainly deploy efforts to categorize an aerial image into one scene-level label, while in real-world scenarios, there often exist multiple scenes in a single image. Therefore, in this paper, we propose to take a step forward to a more practical and challenging task, namely multi-scene recognition in single images. Moreover, we note that manually yielding annotations for such a task is extraordinarily time- and labor-consuming. To address this, we propose a prototype-based memory network to recognize multiple scenes in a single image by leveraging massive well-annotated single-scene images. The proposed network consists of three key components: 1) a prototype learning module, 2) a prototype-inhabiting external memory, and 3) a multi-head attention-based memory retrieval module. To be more specific, we first learn the prototype representation of each aerial scene from single-scene aerial image datasets and store it in an external memory. Afterwards, a multi-head attention-based memory retrieval module is devised to retrieve scene prototypes relevant to query multi-scene images for final predictions. Notably, only a limited number of annotated multi-scene images are needed in the training phase. To facilitate the progress of aerial scene recognition, we produce a new multi-scene aerial image (MAI) dataset. Experimental results on variant dataset configurations demonstrate the effectiveness of our network. Our dataset and codes are publicly available.

Automatic skin lesion classification based on mid-level feature learning.

Dermoscopic images are widely used for melanoma detection. Many existing works based on traditional classification methods and deep learning models have been proposed for automatic skin lesion analysis. The traditional classification methods use hand-crafted features as input. However, due to the strong visual similarity between different classes of skin lesions and complex skin conditions, the hand-crafted features are not discriminative enough and fail in many cases. Recently, deep convolutional neural networks (CNN) have gained popularity since they can automatically learn optimal features during the training phase. Different from existing works, a novel mid-level feature learning method for skin lesion classification task is proposed in this paper. In this method, skin lesion segmentation is first performed to detect the regions of interest (ROI) of skin lesion images. Next, pretrained neural networks including ResNet and DenseNet are used as the feature extractors for the ROI images. Instead of using the extracted features directly as input of classifiers, the proposed method obtains the mid-level feature representations by utilizing the relationships among different image samples based on distance metric learning. The learned feature representation is a soft discriminative descriptor, having more tolerance to the hard samples and hence is more robust to the large intra-class difference and inter-class similarity. Experimental results demonstrate advantages of the proposed mid-level features, and the proposed method obtains state-of-the-art performance compared with the existing CNN based methods.

Local climate zone-based urban land cover classification from multi-seasonal Sentinel-2 images with a recurrent residual network.

The local climate zone (LCZ) scheme was originally proposed to provide an interdisciplinary taxonomy for urban heat island (UHI) studies. In recent years, the scheme has also become a starting point for the development of higher-level products, as the LCZ classes can help provide a generalized understanding of urban structures and land uses. LCZ mapping can therefore theoretically aid in fostering a better understanding of spatio-temporal dynamics of cities on a global scale. However, reliable LCZ maps are not yet available globally. As a first step toward automatic LCZ mapping, this work focuses on LCZ-derived land cover classification, using multi-seasonal Sentinel-2 images. We propose a recurrent residual network (Re-ResNet) architecture that is capable of learning a joint spectral-spatial-temporal feature representation within a unitized framework. To this end, a residual convolutional neural network (ResNet) and a recurrent neural network (RNN) are combined into one end-to-end architecture. The ResNet is able to learn rich spectral-spatial feature representations from single-seasonal imagery, while the RNN can effectively analyze temporal dependencies of multi-seasonal imagery. Cross validations were carried out on a diverse dataset covering seven distinct European cities, and a quantitative analysis of the experimental results revealed that the combined use of the multi-temporal information and Re-ResNet results in an improvement of approximately 7 percent points in overall accuracy. The proposed framework has the potential to produce consistent-quality urban land cover and LCZ maps on a large scale, to support scientific progress in fields such as urban geography and urban climatology.

Recurrently exploring class-wise attention in a hybrid convolutional and bidirectional LSTM network for multi-label aerial image classification.

Aerial image classification is of great significance in the remote sensing community, and many researches have been conducted over the past few years. Among these studies, most of them focus on categorizing an image into one semantic label, while in the real world, an aerial image is often associated with multiple labels, e.g., multiple object-level labels in our case. Besides, a comprehensive picture of present objects in a given high-resolution aerial image can provide a more in-depth understanding of the studied region. For these reasons, aerial image multi-label classification has been attracting increasing attention. However, one common limitation shared by existing methods in the community is that the co-occurrence relationship of various classes, so-called class dependency, is underexplored and leads to an inconsiderate decision. In this paper, we propose a novel end-to-end network, namely class-wise attention-based convolutional and bidirectional LSTM network (CA-Conv-BiLSTM), for this task. The proposed network consists of three indispensable components: (1) a feature extraction module, (2) a class attention learning layer, and (3) a bidirectional LSTM-based sub-network. Particularly, the feature extraction module is designed for extracting fine-grained semantic feature maps, while the class attention learning layer aims at capturing discriminative class-specific features. As the most important part, the bidirectional LSTM-based sub-network models the underlying class dependency in both directions and produce structured multiple object labels. Experimental results on UCM multi-label dataset and DFC15 multi-label dataset validate the effectiveness of our model quantitatively and qualitatively.

Using nano-attapulgite clay compounded hydrophilic urethane foams (AT/HUFs) as biofilm support enhances oil-refinery wastewater treatment in a biofilm membrane bioreactor.

Petroleum refinery wastewater (PRW) treatments based on biofilm membrane bioreactor (BF-MBR) technology is an ideal approach and biofilm supporting material is a critical factor. In this study, BF-MBR with nano-attapulgite clay compounded hydrophilic urethane foams (AT/HUFs) as a biofilm support was used to treat PRW with a hydraulic retention time of 5 h. The removal rate of 500 mg/L chemical oxygen demand (COD), 15 mg/L NH4+ and 180 NTU of turbidity were 99.73%, 97.48% and 99.99%, which were 23%, 20%, and 6% higher than in the control bioreactor, respectively. These results were comparatively higher than that observed for the sequencing batch reactor (SBR). The death rate of the Spirodela polyrrhiza (L.) irrigated with BF-MBR-treated water was 4.44%, which was similar to that of the plants irrigated with tap water (3.33%) and SBR-treated water (5.56%), but significantly lower than that irrigated with raw water (84.44%). The counts demonstrated by qPCR for total bacteria, denitrifiers, nitrite oxidizing bacteria, ammonia oxidizing bacteria, and ammonia-oxidizing archaea were also higher in BF-MBR than those obtained by SBR. Moreover, the results of 16 s rRNA sequencing have demonstrated that the wastewater remediation microbes were enriched in AT/HUFs, e.g., Acidovorax can degrade polycyclic aromatic hydrocarbons, and Sulfuritalea is an efficient nitrite degrader. In summary, BF-MBR using AT/HUF as a biofilm support improves microbiome of the actived sludge and is reliable for oil-refinery wastewater treatment.

Long-term and high-concentration heavy-metal contamination strongly influences the microbiome and functional genes in Yellow River sediments.

The world is facing a hard battle against soil pollution such as heavy metals. Metagenome sequencing, 16S rRNA sequencing, and quantitative polymerase chain reaction (qPCR) were used to examine microbial adaptation mechanism to contaminated sediments under natural conditions. Results showed that sediment from a tributary of the Yellow River, which was named Dongdagou River (DDG) supported less bacterial biomass and owned lower richness than sediment from Maqu (MQ), an uncontaminated site in the upper reaches of the Yellow River. Additionally, microbiome structures in these two sites were different. Metagenome sequencing and functional gene annotations revealed that sediment from DDG contains a larger number of genes related to DNA recombination, DNA damage repair, and heavy-metal resistance. KEGG pathway analysis indicated that the sediment of DDG contains a greater number of enzymes associated with heavy-metal resistance and reduction. Additionally, the bacterial phyla Proteobacteria, Bacteroidetes, and Firmicutes, which harbored a larger suite of metal-resistance genes, were found to be the core functional phyla in the contaminated sediments. Furthermore, sediment in DDG owned higher viral abundance, indicating virus-mediated heavy-metal resistance gene transfer might be an adaptation mechanism. In conclusion, microbiome of sediment from DDG has evolved into an integrated system resistant to long-term heavy-metal pollution.

Scene recognition by manifold regularized deep learning architecture.

Scene recognition is an important problem in the field of computer vision, because it helps to narrow the gap between the computer and the human beings on scene understanding. Semantic modeling is a popular technique used to fill the semantic gap in scene recognition. However, most of the semantic modeling approaches learn shallow, one-layer representations for scene recognition, while ignoring the structural information related between images, often resulting in poor performance. Modeled after our own human visual system, as it is intended to inherit humanlike judgment, a manifold regularized deep architecture is proposed for scene recognition. The proposed deep architecture exploits the structural information of the data, making for a mapping between visible layer and hidden layer. By the proposed approach, a deep architecture could be designed to learn the high-level features for scene recognition in an unsupervised fashion. Experiments on standard data sets show that our method outperforms the state-of-the-art used for scene recognition.

Semi-Supervised Multitask Learning for Scene Recognition.

Scene recognition has been widely studied to understand visual information from the level of objects and their relationships. Toward scene recognition, many methods have been proposed. They, however, encounter difficulty to improve the accuracy, mainly due to two limitations: 1) lack of analysis of intrinsic relationships across different scales, say, the initial input and its down-sampled versions and 2) existence of redundant features. This paper develops a semi-supervised learning mechanism to reduce the above two limitations. To address the first limitation, we propose a multitask model to integrate scene images of different resolutions. For the second limitation, we build a model of sparse feature selection-based manifold regularization (SFSMR) to select the optimal information and preserve the underlying manifold structure of data. SFSMR coordinates the advantages of sparse feature selection and manifold regulation. Finally, we link the multitask model and SFSMR, and propose the semi-supervised learning method to reduce the two limitations. Experimental results report the improvements of the accuracy in scene recognition.

Scene Parsing From an MAP Perspective.

Scene parsing is an important problem in the field of computer vision. Though many existing scene parsing approaches have obtained encouraging results, they fail to overcome within-category inconsistency and intercategory similarity of superpixels. To reduce the aforementioned problem, a novel method is proposed in this paper. The proposed approach consists of three main steps: 1) posterior category probability density function (PDF) is learned by an efficient low-rank representation classifier (LRRC); 2) prior contextual constraint PDF on the map of pixel categories is learned by Markov random fields; and 3) final parsing results are yielded up to the maximum a posterior process based on the two learned PDFs. In this case, the nature of being both dense for within-category affinities and almost zeros for intercategory affinities is integrated into our approach by using LRRC to model the posterior category PDF. Meanwhile, the contextual priori generated by modeling the prior contextual constraint PDF helps to promote the performance of scene parsing. Experiments on benchmark datasets show that the proposed approach outperforms the state-of-the-art approaches for scene parsing.