Search

Background: The advances in deep learning-based pathological image analysis have invoked tremendous insights into cancer prognostication. Still, lack of interpretability remains a significant barrier to clinical application., Methods: We established an integrative prognostic neural network for intrahepatic cholangiocarcinoma (iCCA), towards a comprehensive evaluation of both architectural and fine-grained information from whole-slide images. Then, leveraging on multi-modal data, we conducted extensive interrogative approaches to the models, to extract and visualize the morphological features that most correlated with clinical outcome and underlying molecular alterations., Results: The models were developed and optimized on 373 iCCA patients from our center and demonstrated consistent accuracy and robustness on both internal (n = 213) and external (n = 168) cohorts. The occlusion sensitivity map revealed that the distribution of tertiary lymphoid structures, the geometric traits of the invasive margin, the relative composition of tumor parenchyma and stroma, the extent of necrosis, the presence of the disseminated foci, and the tumor-adjacent micro-vessels were the determining architectural features that impacted on prognosis. Quantifiable morphological vector extracted by CellProfiler demonstrated that tumor nuclei from high-risk patients exhibited significant larger size, more distorted shape, with less prominent nuclear envelope and textural contrast. The multi-omics data (n = 187) further revealed key molecular alterations left morphological imprints that could be attended by the network, including glycolysis, hypoxia, apical junction, mTORC1 signaling, and immune infiltration., Conclusions: We proposed an interpretable deep-learning framework to gain insights into the biological behavior of iCCA. Most of the significant morphological prognosticators perceived by the network are comprehensible to human minds., (© 2024. The Author(s).)

Background Lumbar spine MRI studies are widely used for back pain assessment. Interpretation involves grading lumbar spinal stenosis, which is repetitive and time consuming. Deep learning (DL) could provide faster and more consistent interpretation. Purpose To assess the speed and interobserver agreement of radiologists for reporting lumbar spinal stenosis with and without DL assistance. Materials and Methods In this retrospective study, a DL model designed to assist radiologists in the interpretation of spinal canal, lateral recess, and neural foraminal stenoses on lumbar spine MRI scans was used. Randomly selected lumbar spine MRI studies obtained in patients with back pain who were 18 years and older over a 3-year period, from September 2015 to September 2018, were included in an internal test data set. Studies with instrumentation and scoliosis were excluded. Eight radiologists, each with 2-13 years of experience in spine MRI interpretation, reviewed studies with and without DL model assistance with a 1-month washout period. Time to diagnosis (in seconds) and interobserver agreement (using Gwet κ) were assessed for stenosis grading for each radiologist with and without the DL model and compared with test data set labels provided by an external musculoskeletal radiologist (with 32 years of experience) as the reference standard. Results Overall, 444 images in 25 patients (mean age, 51 years ± 20 [SD]; 14 women) were evaluated in a test data set. DL-assisted radiologists had a reduced interpretation time per spine MRI study, from a mean of 124-274 seconds (SD, 25-88 seconds) to 47-71 seconds (SD, 24-29 seconds) ( P < .001). DL-assisted radiologists had either superior or equivalent interobserver agreement for all stenosis gradings compared with unassisted radiologists. DL-assisted general and in-training radiologists improved their interobserver agreement for four-class neural foraminal stenosis, with κ values of 0.71 and 0.70 (with DL) versus 0.39 and 0.39 (without DL), respectively (both P < .001). Conclusion Radiologists who were assisted by deep learning for interpretation of lumbar spinal stenosis on MRI scans showed a marked reduction in reporting time and superior or equivalent interobserver agreement for all stenosis gradings compared with radiologists who were unassisted by deep learning. © RSNA, 2022 Online supplemental material is available for this article . See also the editorial by Hayashi in this issue.

Background: Inadequate bowel preparation is associated with a decrease in adenoma detection rate (ADR). A deep learning-based bowel preparation assessment system based on the Boston bowel preparation scale (BBPS) has been previously established to calculate the automatic BBPS (e-BBPS) score (ranging 0-20). The aims of this study were to investigate whether there was a statistically inverse relationship between the e-BBPS score and the ADR, and to determine the threshold of e-BBPS score for adequate bowel preparation in colonoscopy screening., Methods: In this prospective, observational study, we trained and internally validated the e-BBPS system using retrospective colonoscopy images and videos from the Endoscopy Center of Wuhan University, annotated by endoscopists. We externally validated the system using colonoscopy images and videos from the First People's Hospital of Yichang and the Third Hospital of Wuhan. To prospectively validate the system, we recruited consecutive patients at Renmin Hospital of Wuhan University aged between 18 and 75 years undergoing colonoscopy. The exclusion criteria included: contraindication to colonoscopy, family polyposis syndrome, inflammatory bowel disease, history of surgery for colorectal or colorectal cancer, known or suspected bowel obstruction or perforation, patients who were pregnant or lactating, inability to receive caecal intubation, and lumen obstruction. We did colonoscopy procedures and collected withdrawal videos, which were reviewed and the e-BBPS system was applied to all colon segments. The primary outcome of this study was ADR, defined as the proportion of patients with one or more conventional adenomas detected during colonoscopy. We calculated the ADR of each e-BBPS score and did a correlation analysis using Spearman analysis., Findings: From May 11 to Aug 10, 2020, 616 patients underwent screening colonoscopies, which evaluated. There was a significant inverse correlation between the e-BBPS score and ADR (Spearman's rank -0·976, p<0·010). The ADR for the e-BBPS scores 1-8 was 28·57%, 28·68%, 26·79%, 19·19%, 17·57%, 17·07%, 14·81%, and 0%, respectively. According to the 25% ADR standard for screening colonoscopy, an e-BBPS score of 3 was set as a threshold to guarantee an ADR of more than 25%, and so high-quality endoscopy. Patients with scores of more than 3 had a significantly lower ADR than those with a score of 3 or less (ADR 15·93% vs 28·03%, p<0·001, 95% CI 0·28-0·66, odds ratio 0·43)., Interpretation: The e-BBPS system has potential to provide a more objective and refined threshold for the quantification of adequate bowel preparation., Funding: Project of Hubei Provincial Clinical Research Center for Digestive Disease Minimally Invasive Incision and Hubei Province Major Science and Technology Innovation Project., Competing Interests: Declaration of interests BZ, CL, and SH are employees of Wuhan EndoAngel Medical Technology Company; the company had no role in this study. All other authors declare no competing interests., (Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Published by Elsevier Ltd.. All rights reserved.)

Background and Aims: EUS is considered one of the most sensitive modalities for pancreatic cancer detection, but it is highly operator-dependent and the learning curve is steep. In this study, we constructed a system named BP MASTER (pancreaticobiliary master) for EUS training and quality control., Methods: The standard procedure of pancreatic EUS was divided into 6 stations. We developed a station classification model and a pancreas/abdominal aorta/portal confluence segmentation model with 19,486 images and 2207 images, respectively. Then, we used 1920 images and 700 images for classification and segmentation internal validation, respectively. To test station recognition we used 396 videos clips. An independent data set containing 180 images was applied for comparing the performance between models and EUS experts. Seven hundred sixty-eight images from 2 other hospitals were used for external validation. A crossover study was conducted to test the system effect on reducing difficulty in ultrasonographics interpretation among trainees., Results: The models achieved 94.2% accuracy in station classification and .836 dice in segmentation at internal validation. At external validation, the models achieved 82.4% accuracy in station classification and .715 dice in segmentation. For the video test, the station classification model achieved a per-frame accuracy of 86.2%. Compared with EUS experts, the models achieved 90.0% accuracy in classification and .77 and .813 dice in blood vessel and pancreas segmentation, which is comparable with that of experts. In the crossover study, trainee station recognition accuracy improved from 67.2% to 78.4% (95% confidence interval, .058-1.663; P < .01)., Conclusions: The BP MASTER system has the potential to play an important role in shortening the pancreatic EUS learning curve and improving EUS quality control in the future., (Copyright © 2020 American Society for Gastrointestinal Endoscopy. Published by Elsevier Inc. All rights reserved.)

Background and Aims: Few artificial intelligence-based technologies have been developed to improve the efficiency of screening for esophageal squamous cell carcinoma (ESCC). Here, we developed and validated a novel system of computer-aided detection (CAD) using a deep neural network (DNN) to localize and identify early ESCC under conventional endoscopic white-light imaging., Methods: We collected 2428 (1332 abnormal, 1096 normal) esophagoscopic images from 746 patients to set up a novel DNN-CAD system in 2 centers and prepared a validation dataset containing 187 images from 52 patients. Sixteen endoscopists (senior, mid-level, and junior) were asked to review the images of the validation set. The diagnostic results, including accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), were compared between the DNN-CAD system and endoscopists., Results: The receiver operating characteristic curve for DNN-CAD showed that the area under the curve was >96%. For the validation dataset, DNN-CAD had a sensitivity, specificity, accuracy, PPV, and NPV of 97.8%, 85.4%, 91.4%, 86.4%, and 97.6%, respectively. The senior group achieved an average diagnostic accuracy of 88.8%, whereas the junior group had a lower value of 77.2%. After referring to the results of DNN-CAD, the average diagnostic ability of the endoscopists improved, especially in terms of sensitivity (74.2% vs 89.2%), accuracy (81.7% vs 91.1%), and NPV (79.3% vs 90.4%)., Conclusions: The novel DNN-CAD system used for screening of early ESCC has high accuracy and sensitivity, and can help endoscopists to detect lesions previously ignored under white-light imaging., (Copyright © 2019 American Society for Gastrointestinal Endoscopy. Published by Elsevier Inc. All rights reserved.)

Copy number variations (CNVs) play an important role in many types of cancer. With the rapid development of next generation sequencing (NGS) techniques, many methods for detecting CNVs of a single sample have emerged: (i) require genome-wide data of both case and control samples, (ii) depend on sequencing depth and GC content correction algorithm, (iii) rely on statistical models built on CNV positive and negative sample datasets. These make them costly in the data analysis and ineffective in the targeted sequencing data. In this study, we developed a novel alignment-free method called DL-CNV to call CNV from the target sequencing data of a single sample. Specifically, we collected two sets of samples. The first set consists of 1301 samples, in which 272 have CNVs in ERBB2 and the second set is composed of 1148 samples with 63 samples containing CNVs in MET. Finally, we found that a testing AUC of 0.9454 for ERBB2 and 0.9220 for MET. Furthermore, we hope to make the CNV detection could be more accurate with clinical "gold standard" (e.g. FISH) information and provide a new research direction, which can be used as the supplement to the existing NGS methods.

The existing deep-learning-based hyperspectral anomaly detection methods detect anomalies by reconstructing a clean background. However, these methods model the background of the hyperspectral image (HSI) through global features, neglecting local features. In complex background scenarios, these methods struggle to obtain accurate background priors for training constraints, thereby limiting the anomaly detection performance. To enhance the capability of the network in extracting local features and improve anomaly detection performance, a hyperspectral anomaly detection method based on differential network is proposed. First, we posit that anomalous pixels are challenging to be reconstructed through the features of surrounding pixels. A differential convolution method is introduced to extract local punctured neighborhood features in the HSI. The differential convolution contains two types of kernels with different receptive fields. These kernels are adopted to obtain the outer window features and inner window features. Second, to improve the feature extraction capability of the network, a local detail attention and a local Transformer attention are proposed. These attention modules enhance the inner window features. Third, the obtained inner window features are subtracted from the outer window features to derive differential features, which encapsulate local punctured neighborhood characteristics. The obtained differential features are employed to reconstruct the background of the HSI. Finally, the anomaly detection results are extracted from the difference between the input HSI and the reconstructed background of the HSI. In the proposed method, for each receptive field kernel, the optimization objective is to reconstruct the input HSI rather than the background HSI. This way circumvents problems where the background constraint biases might affect detection performance. The proposed method offers researchers a new and effective approach for applying deep learning in a local area to the field of hyperspectral anomaly detection. The experiments are conducted with multiple metrics on five real-world datasets. The proposed method outperforms eight state-of-the-art methods in both subjective and objective evaluations. [ABSTRACT FROM AUTHOR]

Skin diseases affect one-third of the global population, posing a major healthcare burden. Deep learning may optimise healthcare workflows through processing skin images via neural networks to make predictions. A focus of deep learning research is skin lesion triage to detect cancer, but this may not translate to the wider scope of >2000 other skin diseases. We searched for studies applying deep learning to skin images, excluding benign/malignant lesions (1/1/2000-23/6/2022, PROSPERO CRD42022309935). The primary outcome was accuracy of deep learning algorithms in disease diagnosis or severity assessment. We modified QUADAS-2 for quality assessment. Of 13,857 references identified, 64 were included. The most studied diseases were acne, psoriasis, eczema, rosacea, vitiligo, urticaria. Deep learning algorithms had high specificity and variable sensitivity in diagnosing these conditions. Accuracy of algorithms in diagnosing acne (median 94%, IQR 86–98; n = 11), rosacea (94%, 90–97; n = 4), eczema (93%, 90–99; n = 9) and psoriasis (89%, 78–92; n = 8) was high. Accuracy for grading severity was highest for psoriasis (range 93–100%, n = 2), eczema (88%, n = 1), and acne (67–86%, n = 4). However, 59 (92%) studies had high risk-of-bias judgements and 62 (97%) had high-level applicability concerns. Only 12 (19%) reported participant ethnicity/skin type. Twenty-four (37.5%) evaluated the algorithm in an independent dataset, clinical setting or prospectively. These data indicate potential of deep learning image analysis in diagnosing and monitoring common skin diseases. Current research has important methodological/reporting limitations. Real-world, prospectively-acquired image datasets with external validation/testing will advance deep learning beyond the current experimental phase towards clinically-useful tools to mitigate rising health and cost impacts of skin disease. [ABSTRACT FROM AUTHOR]

Cartoonization as a special type of artistic style transfer is a difficult image processing task. The current existing artistic style transfer methods cannot generate satisfactory cartoon-style images due to that artistic style images often have delicate strokes and rich hierarchical color changes while cartoon-style images have smooth surfaces without obvious color changes, and sharp edges. To this end, we propose a cartoon loss based generative adversarial network (CartoonLossGAN) for cartoonization. Particularly, we first reuse the encoder part of the discriminator to build a compact generative adversarial network (GAN) based cartoonization architecture. Then we propose a novel cartoon loss function for the architecture. It can imitate the process of sketching to learn the smooth surface of the cartoon image, and imitate the coloring process to learn the coloring of the cartoon image. Furthermore, we also propose an initialization strategy, which is used in the scenario of reusing the discriminator to make our model training easier and more stable. Extensive experimental results demonstrate that our proposed CartoonLossGAN can generate fantastic cartoon-style images, and outperforms four representative methods. [ABSTRACT FROM AUTHOR]

In a colonoscopy, accurate computer-aided polyp detection and segmentation can help endoscopists to remove abnormal tissue. This reduces the chance of polyps developing into cancer, which is of great importance. In this paper, we propose a neural network (parallel residual atrous pyramid network or PRAPNet) based on a parallel residual atrous pyramid module for the segmentation of intestinal polyp detection. We made full use of the global contextual information of the different regions by the proposed parallel residual atrous pyramid module. The experimental results showed that our proposed global prior module could effectively achieve better segmentation results in the intestinal polyp segmentation task compared with the previously published results. The mean intersection over union and dice coefficient of the model in the Kvasir-SEG dataset were 90.4% and 94.2%, respectively. The experimental results outperformed the scores achieved by the seven classical segmentation network models (U-Net, U-Net++, ResUNet++, praNet, CaraNet, SFFormer-L, TransFuse-L). [ABSTRACT FROM AUTHOR]

With the growing popularity of web services, more and more developers are composing multiple services into mashups. Developers show an increasing interest in non-popular services (i.e., long-tail ones), however, there are very scarce studies trying to address the long-tail web service recommendation problem. The major challenges for recommending long-tail services accurately include severe sparsity of historical usage data and unsatisfactory quality of description content. In this paper, we propose to build a deep learning framework to address these challenges and perform accurate long-tail recommendations. To tackle the problem of unsatisfactory quality of description content, we use stacked denoising autoencoders (SDAE) to perform feature extraction. Additionally, we impose the usage records in hot services as a regularization of the encoding output of SDAE, to provide feedback to content extraction. To address the sparsity of historical usage data, we learn the patterns of developers’ preference instead of modeling individual services. Our experimental results on a real-world dataset demonstrate that, with such joint autoencoder based feature representation and content-usage learning framework, the proposed algorithm outperforms the state-of-the-art baselines significantly. [ABSTRACT FROM AUTHOR]

Deep learning-based image super-resolution has shown significantly good performance in improving image quality. In this paper, the RGB-IR cross input and sub-pixel upsampling network is proposed to increase the spatial resolution of an Infrared (IR) image by combining it with a color image of higher spatial resolution obtained with a different imaging modality. Specifically, this is accomplished by fusion of the features map of two RGB-IR inputs in the reconstruction of an infrared image. To improve the accuracy of feature extraction, deconvolution is replaced by sub-pixel convolution to upsample image in the network. Then, the guided filter layer is introduced for image denoising of IR images, and it can preserve the image detail. In addition, the experimental dataset, which is collected by us, contains large numbers of RGB images and corresponding IR images with the same scene. Experimental results on our dataset and other datasets demonstrate that the method is superior to existing methods in accuracy and visual improvement. [ABSTRACT FROM AUTHOR]

• An semi-supervised learning method is proposed to improve diagnostic efficiency of covid-19 CT images. • Based on MixMatch, new data enhancement methods and training methods are introduced to reduce the risk of model over fitting. • Data enhancement method with MixMatch is introduced to reduce model over fitting. • CNN with attention mechanism is constructed to learn features from images more effectively. COVID-19 has spread rapidly all over the world and has infected more than 200 countries and regions. Early screening of suspected infected patients is essential for preventing and combating COVID-19. Computed Tomography (CT) is a fast and efficient tool which can quickly provide chest scan results. To reduce the burden on doctors of reading CTs, in this article, a high precision diagnosis algorithm of COVID-19 from chest CTs is designed for intelligent diagnosis. A semi-supervised learning approach is developed to solve the problem when only small amount of labelled data is available. While following the MixMatch rules to conduct sophisticated data augmentation, we introduce a model training technique to reduce the risk of model over-fitting. At the same time, a new data enhancement method is proposed to modify the regularization term in MixMatch. To further enhance the generalization of the model, a convolutional neural network based on an attention mechanism is then developed that enables to extract multi-scale features on CT scans. The proposed algorithm is evaluated on an independent CT dataset of the chest from COVID-19 and achieves the area under the receiver operating characteristic curve (AUC) value of 0.932, accuracy of 90.1%, sensitivity of 91.4%, specificity of 88.9%, and F1-score of 89.9%. The results show that the proposed algorithm can accurately diagnose whether a chest CT belongs to a positive or negative indication of COVID-19, and can help doctors to diagnose rapidly in the early stages of a COVID-19 outbreak. [ABSTRACT FROM AUTHOR]

• A novel characteristics imaging model of ghost imaging is proposed. • We use the pretrained reference beams for target recognition in our imaging scheme. • Influencing factors in real application sceneries are considered. • The target of interest in real imaging area is recognized. • The occluded target can be well recognized. It is important to recognize the target of interest in an imaging area. Target recognition is applicable in fields such as driverless cars and military technology, while target recognition precision is influenced by environmental disturbance. Ghost imaging (GI) presents an anti-disturbance property with potential in target recognition. This paper proposes a characteristic imaging model for target recognition based on preprocessing in computational GI with deep learning. Different from GI trained by detected signals, reference beams pretrained by a deep learning network are used in correlation calculation in our imaging scheme. The imaging object, presenting identical characteristics with the training object, is easier to recognize. Influencing factors in real application scenes are considered, including the size and posture of the imaging object and occlusions, and the results show that our imaging scheme can recognize the target of interest. These results demonstrate potential applications in target recognition. [ABSTRACT FROM AUTHOR]