Rapid Endoscopic Diagnosis of Benign Ulcerative Colorectal Diseases With an Artificial Intelligence Contextual Framework.

BACKGROUND & AIMS: Benign ulcerative colorectal diseases (UCDs) such as ulcerative colitis, Crohn's disease, ischemic colitis, and intestinal tuberculosis share similar phenotypes with different etiologies and treatment strategies. To accurately diagnose closely related diseases like UCDs, we hypothesize that contextual learning is critical in enhancing the ability of the artificial intelligence models to differentiate the subtle differences in lesions amidst the vastly divergent spatial contexts. METHODS: White-light colonoscopy datasets of patients with confirmed UCDs and healthy controls were retrospectively collected. We developed a Multiclass Contextual Classification (MCC) model that can differentiate among the mentioned UCDs and healthy controls by incorporating the tissue object contexts surrounding the individual lesion region in a scene and spatial information from other endoscopic frames (video-level) into a unified framework. Internal and external datasets were used to validate the model's performance. RESULTS: Training datasets included 762 patients, and the internal and external testing cohorts included 257 patients and 293 patients, respectively. Our MCC model provided a rapid reference diagnosis on internal test sets with a high averaged area under the receiver operating characteristic curve (image-level: 0.950 and video-level: 0.973) and balanced accuracy (image-level: 76.1% and video-level: 80.8%), which was superior to junior endoscopists (accuracy: 71.8%, P < .0001) and similar to experts (accuracy: 79.7%, P = .732). The MCC model achieved an area under the receiver operating characteristic curve of 0.988 and balanced accuracy of 85.8% using external testing datasets. CONCLUSIONS: These results enable this model to fit in the routine endoscopic workflow, and the contextual framework to be adopted for diagnosing other closely related diseases.

Dual Networks Based 3D Multi-Person Pose Estimation From Monocular Video.

Monocular 3D human pose estimation has made progress in recent years. Most of the methods focus on single persons, which estimate the poses in the person-centric coordinates, i.e., the coordinates based on the center of the target person. Hence, these methods are inapplicable for multi-person 3D pose estimation, where the absolute coordinates (e.g., the camera coordinates) are required. Moreover, multi-person pose estimation is more challenging than single pose estimation, due to inter-person occlusion and close human interactions. Existing top-down multi-person methods rely on human detection (i.e., top-down approach), and thus suffer from the detection errors and cannot produce reliable pose estimation in multi-person scenes. Meanwhile, existing bottom-up methods that do not use human detection are not affected by detection errors, but since they process all persons in a scene at once, they are prone to errors, particularly for persons in small scales. To address all these challenges, we propose the integration of top-down and bottom-up approaches to exploit their strengths. Our top-down network estimates human joints from all persons instead of one in an image patch, making it robust to possible erroneous bounding boxes. Our bottom-up network incorporates human-detection based normalized heatmaps, allowing the network to be more robust in handling scale variations. Finally, the estimated 3D poses from the top-down and bottom-up networks are fed into our integration network for final 3D poses. To address the common gaps between training and testing data, we do optimization during the test time, by refining the estimated 3D human poses using high-order temporal constraint, re-projection loss, and bone length regularizations. We also introduce a two-person pose discriminator that enforces natural two-person interactions. Finally, we apply a semi-supervised method to overcome the 3D ground-truth data scarcity. Our evaluations demonstrate the effectiveness of the proposed method and its individual components. Our code and pretrained models are available publicly: https://github.com/3dpose/3D-Multi-Person-Pose.

Learning to Remove Rain in Video With Self-Supervision.

In heavy rain video, rain streak and rain accumulation are the most common causes of degradation. They occlude background information and can significantly impair the visibility. Most existing methods rely heavily on the synthetic training data, and thus raise the domain gap problem that prevents the trained models from performing adequately in real testing cases. Unlike these methods, we introduce a self-learning method to remove both rain streaks and rain accumulation without using any ground-truth clean images in training our model, which consequently can alleviate the domain gap issue. The main idea is based on the assumptions that (1) adjacent clean frames can be aligned or warped from one frame to another frame, (2) rain streaks are distributed randomly in the temporal domain, (3) the rain streak/accumulation related variables/priors can be inferred reliably from the information within the images/sequences. Based on these assumptions, we construct an augmented Self-Learned Deraining Network (SLDNet+) to remove both rain streaks and rain accumulation by utilizing temporal correlation, consistency, and rain-related priors. For the temporal correlation, our SLDNet+ takes rain degraded adjacent frames as its input, aligns them, and learns to predict the clean version of the current frame. For the temporal consistency, a new loss is designed to build a robust mapping between the predicted clean frame and non-rain regions from the adjacent rain frames. For the rain-streak-related prior, the rain streak removal network is optimized jointly with motion estimation and rain region detection; while for the rain-accumulation-related prior, a novel non-local video rain accumulation removal method is developed to estimate the accumulation-lines from the whole input video and to offer better color constancy and temporal smoothness. Extensive experiments show the effectiveness of our approach, which provides superior results compared with the existing state of the art methods both quantitatively and qualitatively. The source code will be made publicly available at: https://github.com/flyywh/CVPR-2020-Self-Rain-Removal-Journal.

Feature-Aligned Video Raindrop Removal With Temporal Constraints.

Existing adherent raindrop removal methods focus on the detection of the raindrop locations, and then use inpainting techniques or generative networks to recover the background behind raindrops. Yet, as adherent raindrops are diverse in sizes and appearances, the detection is challenging for both single image and video. Moreover, unlike rain streaks, adherent raindrops tend to cover the same area in several frames. Addressing these problems, our method employs a two-stage video-based raindrop removal method. The first stage is the single image module, which generates initial clean results. The second stage is the multiple frame module, which further refines the initial results using temporal constraints, namely, by utilizing multiple input frames in our process and applying temporal consistency between adjacent output frames. Our single image module employs a raindrop removal network to generate initial raindrop removal results, and create a mask representing the differences between the input and initial output. Once the masks and initial results for consecutive frames are obtained, our multiple-frame module aligns the frames in both the image and feature levels and then obtains the clean background. Our method initially employs optical flow to align the frames, and then utilizes deformable convolution layers further to achieve feature-level frame alignment. To remove small raindrops and recover correct backgrounds, a target frame is predicted from adjacent frames. A series of unsupervised losses are proposed so that our second stage, which is the video raindrop removal module, can self-learn from video data without ground truths. Experimental results on real videos demonstrate the state-of-art performance of our method both quantitatively and qualitatively.

Recurrent Multi-Frame Deraining: Combining Physics Guidance and Adversarial Learning.

Existing video rain removal methods mainly focus on rain streak removal and are solely trained based on the synthetic data, which neglect more complex degradation factors, e.g., rain accumulation, and the prior knowledge in real rain data. Thus, in this paper, we build a more comprehensive rain model with several degradation factors and construct a novel two-stage video rain removal method that combines the power of synthetic videos and real data. Specifically, a novel two-stage progressive network is proposed: recovery guided by a physics model, and further restoration by adversarial learning. The first stage performs an inverse recovery process guided by our proposed rain model. An initially estimated background frame is obtained based on the input rain frame. The second stage employs adversarial learning to refine the result, i.e., recovering the overall color and illumination distributions of the frame, the background details that are failed to be recovered in the first stage, and removing the artifacts generated in the first stage. Furthermore, we also introduce a more comprehensive rain model that includes degradation factors, e.g., occlusion and rain accumulation, which appear in real scenes yet ignored by existing methods. This model, which generates more realistic rain images, will train and evaluate our models better. Extensive evaluations on synthetic and real videos show the effectiveness of our method in comparisons to the state-of-the-art methods. Our datasets, results and code are available at: https://github.com/flyywh/Recurrent-Multi-Frame-Deraining.

Artificial intelligence-enhanced white-light colonoscopy with attention guidance predicts colorectal cancer invasion depth.

BACKGROUND AND AIMS: Endoscopic submucosal dissection (ESD) and EMR are applied in treating superficial colorectal neoplasms but are contraindicated by deeply invasive colorectal cancer (CRC). The invasion depth of neoplasms can be examined by an automated artificial intelligence (AI) system to determine the applicability of ESD and EMR. METHODS: A deep convolutional neural network with a tumor localization branch to guide invasion depth classification was constructed on the GoogLeNet architecture. The model was trained using 7734 nonmagnified white-light colonoscopy (WLC) images supplemented by image augmentation from 657 lesions labeled with histopathologic analysis of invasion depth. An independent testing dataset consisting of 1634 WLC images from 156 lesions was used to validate the model. RESULTS: For predicting noninvasive and superficially invasive neoplasms, the model achieved an overall accuracy of 91.1% (95% confidence interval [CI], 89.6%-92.4%), with 91.2% sensitivity (95% CI, 88.8%-93.3%) and 91.0% specificity (95% CI, 89.0%-92.7%) at an optimal cutoff of .41 and the area under the receiver operating characteristic (AUROC) curve of .970 (95% CI, .962-.978). Inclusion of the advanced CRC data significantly increased the sensitivity in differentiating superficial neoplasms from deeply invasive early CRC to 65.3% (95% CI, 61.9%-68.8%) with an AUROC curve of .729 (95% CI, .699-.759), similar to experienced endoscopists (.691; 95% CI, .624-.758). CONCLUSIONS: We have developed an AI-enhanced attention-guided WLC system that differentiates noninvasive or superficially submucosal invasive neoplasms from deeply invasive CRC with high accuracy, sensitivity, and specificity.

Single Image Deraining: From Model-Based to Data-Driven and Beyond.

The goal of single-image deraining is to restore the rain-free background scenes of an image degraded by rain streaks and rain accumulation. The early single-image deraining methods employ a cost function, where various priors are developed to represent the properties of rain and background layers. Since 2017, single-image deraining methods step into a deep-learning era, and exploit various types of networks, i.e., convolutional neural networks, recurrent neural networks, generative adversarial networks, etc., demonstrating impressive performance. Given the current rapid development, in this paper, we provide a comprehensive survey of deraining methods over the last decade. We summarize the rain appearance models, and discuss two categories of deraining approaches: model-based and data-driven approaches. For the former, we organize the literature based on their basic models and priors. For the latter, we discuss the developed ideas related to architectures, constraints, loss functions, and training datasets. We present milestones of single-image deraining methods, review a broad selection of previous works in different categories, and provide insights on the historical development route from the model-based to data-driven methods. We also summarize performance comparisons quantitatively and qualitatively. Beyond discussing the technicality of deraining methods, we also discuss the future possible directions.

Joint Rain Detection and Removal from a Single Image with Contextualized Deep Networks.

Rain streaks, particularly in heavy rain, not only degrade visibility but also make many computer vision algorithms fail to function properly. In this paper, we address this visibility problem by focusing on single-image rain removal, even in the presence of dense rain streaks and rain-streak accumulation, which is visually similar to mist or fog. To achieve this, we introduce a new rain model and a deep learning architecture. Our rain model incorporates a binary rain map indicating rain-streak regions, and accommodates various shapes, directions, and sizes of overlapping rain streaks, as well as rain accumulation, to model heavy rain. Based on this model, we construct a multi-task deep network, which jointly learns three targets: the binary rain-streak map, rain streak layers, and clean background, which is our ultimate output. To generate features that can be invariant to rain steaks, we introduce a contextual dilated network, which is able to exploit regional contextual information. To handle various shapes and directions of overlapping rain streaks, our strategy is to utilize a recurrent process that progressively removes rain streaks. Our binary map provides a constraint and thus additional information to train our network. Extensive evaluation on real images, particularly in heavy rain, shows the effectiveness of our model and architecture.

Convolutional Neural Networks for Automated Fracture Detection and Localization on Wrist Radiographs.

PURPOSE: To demonstrate the feasibility and performance of an object detection convolutional neural network (CNN) for fracture detection and localization on wrist radiographs. MATERIALS AND METHODS: Institutional review board approval was obtained with waiver of consent for this retrospective study. A total of 7356 wrist radiographic studies were extracted from a hospital picture archiving and communication system. Radiologists annotated all radius and ulna fractures with bounding boxes. The dataset was split into training (90%) and validation (10%) sets and used to train fracture localization models for frontal and lateral images. Inception-ResNet Faster R-CNN architecture was implemented as a deep learning model. The models were tested on an unseen test set of 524 consecutive emergency department wrist radiographic studies with two radiologists in consensus as the reference standard. Per-fracture, per-image (ie, per-view), and per-study sensitivity and specificity were determined. Area under the receiver operating characteristic curve (AUC) analysis was performed. RESULTS: The model detected and correctly localized 310 (91.2%) of 340 and 236 (96.3%) of 245 of all radius and ulna fractures on the frontal and lateral views, respectively. The per-image sensitivity, specificity, and AUC were 95.7% (95% confidence interval [CI]: 92.4%, 97.8%), 82.5% (95% CI: 77.4%, 86.8%), and 0.918 (95% CI: 0.894, 0.941), respectively, for the frontal view and 96.7% (95% CI: 93.6%, 98.6%), 86.4% (95% CI: 81.9%, 90.2%), and 0.933 (95% CI: 0.912, 0.954), respectively, for the lateral view. The per-study sensitivity, specificity, and AUC were 98.1% (95% CI: 95.6%, 99.4%), 72.9% (95% CI: 67.1%, 78.2%), and 0.895 (95% CI: 0.870, 0.920), respectively. CONCLUSION: The ability of an object detection CNN to detect and localize radius and ulna fractures on wrist radiographs with high sensitivity and specificity was demonstrated.© RSNA, 2019.

Single Image Rain Streak Decomposition Using Layer Priors.

Rain streaks impair visibility of an image and introduce undesirable interference that can severely affect the performance of computer vision and image analysis systems. Rain streak removal algorithms try to recover a rain streak free background scene. In this paper, we address the problem of rain streak removal from a single image by formulating it as a layer decomposition problem, with a rain streak layer superimposed on a background layer containing the true scene content. Existing decomposition methods that address this problem employ either sparse dictionary learning methods or impose a low rank structure on the appearance of the rain streaks. While these methods can improve the overall visibility, their performance can often be unsatisfactory, for they tend to either over-smooth the background images or generate -images that still contain noticeable rain streaks. To address the problems, we propose a method that imposes priors for both the background and rain streak layers. These priors are based on Gaussian mixture models learned on small patches that can accommodate a variety of background appearances as well as the appearance of the rain streaks. Moreover, we introduce a structure residue recovery step to further separate the background residues and improve the decomposition quality. Quantitative evaluation shows our method outperforms existing methods by a large margin. We overview our method and demonstrate its effectiveness over prior work on a number of examples.

Adherent Raindrop Modeling, Detectionand Removal in Video.

Raindrops adhered to a windscreen or window glass can significantly degrade the visibility of a scene. Modeling, detecting and removing raindrops will, therefore, benefit many computer vision applications, particularly outdoor surveillance systems and intelligent vehicle systems. In this paper, a method that automatically detects and removes adherent raindrops is introduced. The core idea is to exploit the local spatio-temporal derivatives of raindrops. To accomplish the idea, we first model adherent raindrops using law of physics, and detect raindrops based on these models in combination with motion and intensity temporal derivatives of the input video. Having detected the raindrops, we remove them and restore the images based on an analysis that some areas of raindrops completely occludes the scene, and some other areas occlude only partially. For partially occluding areas, we restore them by retrieving as much as possible information of the scene, namely, by solving a blending function on the detected partially occluding areas using the temporal intensity derivative. For completely occluding areas, we recover them by using a video completion technique. Experimental results using various real videos show the effectiveness of our method.

Separating reflection components of textured surfaces using a single image.

In inhomogeneous objects, highlights are linear combinations of diffuse and specular reflection components. A number of methods have been proposed to separate or decompose these two components. To our knowledge, all methods that use a single input image require explicit color segmentation to deal with multicolored surfaces. Unfortunately, for complex textured images, current color segmentation algorithms are still problematic to segment correctly. Consequently, a method without explicit color segmentation becomes indispensable and this paper presents such a method. The method is based solely on colors, particularly chromaticity, without requiring any geometrical information. One of the basic ideas is to iteratively compare the intensity logarithmic differentiation of an input image and its specular-free image. A specular-free image is an image that has exactly the same geometrical profile as the diffuse component of the input image and that can be generated by shifting each pixel's intensity and maximum chromaticity nonlinearly. Unlike existing methods using a single image, all processes in the proposed method are done locally, involving a maximum of only two neighboring pixels. This local operation is useful for handling textured objects with complex multicolored scenes. Evaluations by comparison with the results of polarizing filters demonstrate the effectiveness of the proposed method.

Color constancy through inverse-intensity chromaticity space.

Existing color constancy methods cannot handle both uniformly colored surfaces and highly textured surfaces in a single integrated framework. Statistics-based methods require many surface colors and become error prone when there are only a few surface colors. In contrast, dichromatic-based methods can successfully handle uniformly colored surfaces but cannot be applied to highly textured surfaces, since they require precise color segmentation. We present a single integrated method to estimate illumination chromaticity from single-colored and multicolored surfaces. Unlike existing dichromatic-based methods, the proposed method requires only rough highlight regions without segmenting the colors inside them. We show that, by analyzing highlights, a direct correlation between illumination chromaticity and image chromaticity can be obtained. This correlation is clearly described in "inverse-intensity chromaticity space," a novel two-dimensional space that we introduce. In addition, when Hough transform and histogram analysis is utilized in this space, illumination chromaticity can be estimated robustly, even for a highly textured surface.

Separating reflection components based on chromaticity and noise analysis.

Many algorithms in computer vision assume diffuse only reflections and deem specular reflections to be outliers. However, in the real world, the presence of specular reflections is inevitable since there are many dielectric inhomogeneous objects which have both diffuse and specular reflections. To resolve this problem, we present a method to separate the two reflection components. The method is principally based on the distribution of specular and diffuse points in a two-dimensional maximum chromaticity-intensity space. We found that, by utilizing the space and known illumination color, the problem of reflection component separation can be simplified into the problem of identifying diffuse maximum chromaticity. To be able to identify the diffuse maximum chromaticity correctly, an analysis of the noise is required since most real images suffer from it. Unlike existing methods, the proposed method can separate the reflection components robustly for any kind of surface roughness and light direction.