Automatic Fracture Detection Research Articles

Comparative Analysis of Neural Network Architectures for Automated Fracture Detection in Hand X-ray Images

The application of several neural network architectures—including Fully Connected Networks (FCN), Convolutional Neural Networks (CNN), pretrained ResNet, Vision Transformer (ViT-B-16)—for the classification of hand X-ray images into "Fractured" and "Not Fractured"—categories is investigated in this work. The main goals are to evaluate these models' fracture detection ability and determine which architectural design fits this work. Because transfer learning let the model use past information from big-scale picture datasets, the pretrained ResNet model emerged as the most effective with high accuracy, stability, and resilience. The bespoke CNN also performed well, displaying excellent feature extraction powers especially for medical imaging. But the non-pretrained ResNet model overfitted, meaning deeper networks find it difficult to generalize without pretraining. Though innovative, the Vision Transformer performed poorly since it depends on a lot of training data and finds difficult learning of intricate spatial properties from little datasets. Although acting as a baseline, the FCN's simple architecture and incapacity to detect spatial hierarchies in images meant it could not match the efficacy of CNN models. Emphasizing the important function of transfer learning in clinical applications, the results show that pretrained CNN architectures, especially ResNet, offer the most consistent and accurate method for automatic fracture diagnosis in medical pictures.

Interpretable Severity Scoring of Pelvic Trauma Through Automated Fracture Detection and Bayesian Inference.

Pelvic ring disruptions result from blunt injury mechanisms and are potentially lethal mainly due to associated injuries and massive pelvic hemorrhage. The severity of pelvic fractures in trauma victims is frequently assessed by grading the fracture according to the Tile AO/OTA classification in whole-body Computed Tomography (CT) scans. Due to the high volume of whole-body CT scans generated in trauma centers, the overall information content of a single whole-body CT scan and low manual CT reading speed, an automatic approach to Tile classification would provide substantial value, e. g., to prioritize the reading sequence of the trauma radiologists or enable them to focus on other major injuries in multi-trauma patients. In such a high-stakes scenario, an automated method for Tile grading should ideally be transparent such that the symbolic information provided by the method follows the same logic a radiologist or orthopedic surgeon would use to determine the fracture grade. This paper introduces an automated yet interpretable pelvic trauma decision support system to assist radiologists in fracture detection and Tile grading. To achieve interpretability despite processing high-dimensional whole-body CT images, we design a neurosymbolic algorithm that operates similarly to human interpretation of CT scans. The algorithm first detects relevant pelvic fractures on CTs with high specificity using Faster-RCNN. To generate robust fracture detections and associated detection (un)certainties, we perform test-time augmentation of the CT scans to apply fracture detection several times in a self-ensembling approach. The fracture detections are interpreted using a structural causal model based on clinical best practices to infer an initial Tile grade. We apply a Bayesian causal model to recover likely co-occurring fractures that may have been rejected initially due to the highly specific operating point of the detector, resulting in an updated list of detected fractures and corresponding final Tile grade. Our method is transparent in that it provides fracture location and types, as well as information on important counterfactuals that would invalidate the system's recommendation. Our approach achieves an AUC of 0.89/0.74 for translational and rotational instability,which is comparable to radiologist performance. Despite being designed for human-machine teaming, our approach does not compromise on performance compared to previous black-box methods.

Robust Multi-View Fracture Detection inthe Presence ofOther Abnormalities Using HAMIL-Net.

Foot and ankle fractures are the most common military health problem. Automated diagnosis can save time and personnel. It is crucial to distinguish fractures not only from normal healthy cases, but also robust against the presence of other orthopedic pathologies. Artificial intelligence (AI) deep learning has been shown to be promising. Previously, we have developed HAMIL-Net to automatically detect orthopedic injuries for upper extremity injuries. In this research, we investigated the performance of HAMIL-Net for detecting foot and ankle fractures in the presence of other abnormalities. HAMIL-Net is a novel deep neural network consisting of a hierarchical attention layer followed by a multiple-instance learning layer. The design allowed it to deal with imaging studies with multiple views. We used 148K musculoskeletal imaging studies for 51K Veterans at VA San Diego in the past 20 years to create datasets for this research. We annotated each study by a semi-automated pipeline leveraging radiology reports written by board-certified radiologists and extracting findings with a natural language processing tool and manually validated the annotations. HAMIL-Net can be trained with study-level, multiple-view examples, and detect foot and ankle fractures with a 0.87 area under the receiver operational curve, but the performance dropped when tested by cases including other abnormalities. By integrating a fracture specialized model with one that detecting a broad range of abnormalities, HAMIL-Net's accuracy of detecting any abnormality improved from 0.53 to 0.77 and F-score from 0.46 to 0.86. We also reported HAMIL-Net's performance under different study types including for young (age 18-35) patients. Automated fracture detection is promising but to be deployed in clinical use, presence of other abnormalities must be considered to deliver its full benefit. Our results with HAMIL-Net showed that considering other abnormalities improved fracture detection and allowed for incidental findings of other musculoskeletal abnormalities pertinent or superimposed on fractures.

Automated fracture detection in the ulna and radius using deep learning on upper extremity radiographs.

This study aimed to detect single or multiple fractures in the ulna or radius using deep learning techniques fed on upper-extremity radiographs. The data set used in the retrospective study consisted of different types of upper extremity radiographs obtained from an open-source dataset, with 4,480 images with fractures and 4,383 images without fractures. All fractures involved the ulna or radius. The proposed method comprises two distinct stages. The initial phase, referred to as preprocessing, involved the removal of radiographic backgrounds, followed by the elimination of nonbone tissue. In the second phase, images consisting only of bone tissue were processed using deep learning models, such as RegNetX006, EfficientNet B0, and InceptionResNetV2. Thus, whether one or more fractures of the ulna or the radius are present was determined. To measure the performance of the proposed method, raw images, images generated by background deletion, and bone tissue removal were classified separately using RegNetX006, EfficientNet B0, and InceptionResNetV2 models. Performance was assessed by accuracy, F1 score, Matthew's correlation coefficient, receiver operating characteristic area under the curve, sensitivity, specificity, and precision using 10-fold cross-validation, which is a widely accepted technique in statistical analysis. The best classification performance was obtained with the proposed preprocessing and RegNetX006 architecture. The values obtained for various metrics were as follows: accuracy (0.9921), F1 score (0.9918), Matthew's correlation coefficient (0.9842), area under the curve (0.9918), sensitivity (0.9974), specificity (0.9863), and precision (0.9923). The proposed preprocessing method is able to detect fractures of the ulna and radius by artificial intelligence.

Automatic fracture detection from the images of electrical image logs using Mask R-CNN

This paper presents a novel approach for automatically detecting fractures from electrical image logs using a state-of-the-art object detection and segmentation method, i.e., the Mask Regional Convolutional Neural Network (Mask R-CNN). First, we apply image processing algorithms of inpainting and augmentation to generate the basic dataset. The fractures in those processed images are properly identified and labeled. The processed images with the annotated fractures will be used as input data for the training and testing of the Mask R-CNN model. Subsequently, the input data are randomly split into training dataset, validation dataset, and testing dataset. The Mask R-CNN model is trained using the training and validation datasets. The performance and accuracy of the trained model is tested by comparing the predicted fractures from the testing dataset to the actual fractures that are identified manually. The testing results demonstrate that the trained Mask R-CNN model has precision and recall values of 96% and 92% in detecting fractures from the image logs. Last, we further validate the proposed approach by comparing the fractures detected using the developed Mask R-CNN model against those identified from three core samples. The comparison also indicates a high accuracy of the Mask R-CNN model.

Classification of Long-Bone Fractures Using Modified Faster RCNN for X-Ray Images

Objectives: Methods in machine learning have been shown to be an essential tool for the diagnosis and treatment of disease. The scientists are constantly looking for new technology that might improve current clinical practise. One of the most well-liked study fields continues to be computerized bone fracture identification and categorization. Furthermore, identifying cracks and determining where they occur have been challenges for historically developed approaches for the long bone fracture method. Methods: We propose a new approach for automatic fracture detection in X-Ray images. This approach is built on top of two-stage fracture detection deep learning algorithm called Faster R-CNN with a major modification of using rotated bounding box. The described procedure is divided into four main steps: X-ray images are used to: I identify the bone contour; (ii) identify fracture-points or fractures; (iii) find a similar set of shapes that are compatible with the identification of cracks; and (iv) classify and thoroughly evaluate the fracture-type using bounding box. Finding: The resolution process uses the stretched numerical conventional line techniques (RDS), arcs, discrete curvature, and shape directory, among other mathematical characteristics of digital curves. We assessed the suggested model’s performance in terms of classification and detection. We divide x-ray pictures of bone fractures into 2 groups, fracture and non-fracture, and we also use a rectangular box to identify the location of fractures. Besides, an additional benefit of rotated bounding box is that it can provide relative information on the orientation and length of fracture without the further segmentation and measurement step. Novelty: This study develops a new approach to automatically and accurately detect fractures in X-Ray images. The proposed approach is developed on the basis of Faster R-CNN algorithm with a major modification for the task of rotated bounding box prediction. Keywords: Long bone; X-Ray images; Classification; Detection; Faster RCNN; Fracture; Non-Fracture; Types

Development and external validation of automated detection, classification, and localization of ankle fractures: inside the black box of a convolutional neural network (CNN)

PurposeConvolutional neural networks (CNNs) are increasingly being developed for automated fracture detection in orthopaedic trauma surgery. Studies to date, however, are limited to providing classification based on the entire image—and only produce heatmaps for approximate fracture localization instead of delineating exact fracture morphology. Therefore, we aimed to answer (1) what is the performance of a CNN that detects, classifies, localizes, and segments an ankle fracture, and (2) would this be externally valid?MethodsThe training set included 326 isolated fibula fractures and 423 non-fracture radiographs. The Detectron2 implementation of the Mask R-CNN was trained with labelled and annotated radiographs. The internal validation (or ‘test set’) and external validation sets consisted of 300 and 334 radiographs, respectively. Consensus agreement between three experienced fellowship-trained trauma surgeons was defined as the ground truth label. Diagnostic accuracy and area under the receiver operator characteristic curve (AUC) were used to assess classification performance. The Intersection over Union (IoU) was used to quantify accuracy of the segmentation predictions by the CNN, where a value of 0.5 is generally considered an adequate segmentation.ResultsThe final CNN was able to classify fibula fractures according to four classes (Danis-Weber A, B, C and No Fracture) with AUC values ranging from 0.93 to 0.99. Diagnostic accuracy was 89% on the test set with average sensitivity of 89% and specificity of 96%. External validity was 89–90% accurate on a set of radiographs from a different hospital. Accuracies/AUCs observed were 100/0.99 for the ‘No Fracture’ class, 92/0.99 for ‘Weber B’, 88/0.93 for ‘Weber C’, and 76/0.97 for ‘Weber A’. For the fracture bounding box prediction by the CNN, a mean IoU of 0.65 (SD ± 0.16) was observed. The fracture segmentation predictions by the CNN resulted in a mean IoU of 0.47 (SD ± 0.17).ConclusionsThis study presents a look into the ‘black box’ of CNNs and represents the first automated delineation (segmentation) of fracture lines on (ankle) radiographs. The AUC values presented in this paper indicate good discriminatory capability of the CNN and substantiate further study of CNNs in detecting and classifying ankle fractures.Level of evidenceII, Diagnostic imaging study.

A pediatric wrist trauma X-ray dataset (GRAZPEDWRI-DX) for machine learning

Digital radiography is widely available and the standard modality in trauma imaging, often enabling to diagnose pediatric wrist fractures. However, image interpretation requires time-consuming specialized training. Due to astonishing progress in computer vision algorithms, automated fracture detection has become a topic of research interest. This paper presents the GRAZPEDWRI-DX dataset containing annotated pediatric trauma wrist radiographs of 6,091 patients, treated at the Department for Pediatric Surgery of the University Hospital Graz between 2008 and 2018. A total number of 10,643 studies (20,327 images) are made available, typically covering posteroanterior and lateral projections. The dataset is annotated with 74,459 image tags and features 67,771 labeled objects. We de-identified all radiographs and converted the DICOM pixel data to 16-Bit grayscale PNG images. The filenames and the accompanying text files provide basic patient information (age, sex). Several pediatric radiologists annotated dataset images by placing lines, bounding boxes, or polygons to mark pathologies like fractures or periosteal reactions. They also tagged general image characteristics. This dataset is publicly available to encourage computer vision research.

FracDetect: A novel algorithm for 3D fracture detection in digital fractured rocks

Fractures have a governing effect on the physical properties of fractured rocks, such as permeability. Accurate representation of 3D fractures is, therefore, required for precise analysis of digital fractured rocks. However, conventional segmentation methods fail to detect and label the fractures with aperture sizes near or below the resolution of 3D micro-computed tomographic (micro-CT) images, which are visible in the greyscale images, and where greyscale intensity convolution between different phases exists. In addition, conventional methods are highly subjective to user interpretation. Herein, a novel algorithm for the automatic detection of fractures from greyscale 3D micro-CT images is proposed. The algorithm involves a low-level early vision stage, which identifies potential fractures, followed by a high-level interpretative stage, which enforces planar continuity to reject false positives and more reliably extract planar fractures from digital rock images. A manually segmented fractured shale sample was used as the groundtruth, with which the efficacy of the algorithm in 3D fracture detection was validated. Following this, the proposed and conventional methods were applied to detect fractures in digital fractured coal and shale samples. Based on these analyses, the impact of fracture detection accuracy on the analysis of fractured rocks' physical properties was inferred.

An attention‐based cascade R‐CNN model for sternum fracture detection in X‐ray images

Fracture is one of the most common and unexpected traumas. If not treated in time, it may cause serious consequences such as joint stiffness, traumatic arthritis, and nerve injury. Using computer vision technology to detect fractures can reduce the workload and misdiagnosis of fractures and also improve the fracture detection speed. However, there are still some problems in sternum fracture detection, such as the low detection rate of small and occult fractures. In this work, the authors have constructed a dataset with 1227 labelled X-ray images for sternum fracture detection. The authors designed a fully automatic fracture detection model based on a deep convolution neural network (CNN). The authors used cascade R-CNN, attention mechanism, and atrous convolution to optimise the detection of small fractures in a large X-ray image with big local variations. The authors compared the detection results of YOLOv5 model, cascade R-CNN and other state-of-the-art models. The authors found that the convolution neural network based on cascade and attention mechanism models has a better detection effect and arrives at an mAP of 0.71, which is much better than using the YOLOv5 model (mAP = 0.44) and cascade R-CNN (mAP = 0.55).

Automatic fracture detection and characterization from unwrapped drill-core images using mask R–CNN

Drill cores provide reliable information about fractures in subsurface formations as they present a clear and direct view of fractures. Core observation and image log interpretation are usually integrated for fracture analysis of underground layers. There has been a strong move towards developing automated fracture detection methods, however, the focus has been on extracting fracture information from log images, such as acoustic or resistivity image logs. Such efforts using core images are significantly less. This paper presents a machine learning-based approach for automatic fracture recognition from unwrapped drill-core images. The proposed method applies a state-of-the-art convolutional neural network for object identification and segmentation. The study also investigates the feasibility of using synthetic fracture images for the training of a learning model. Synthetic data can provide an alternative to real data, and thus address data availability issues common for supervised machine learning applications. We first create two types of synthetic data by using masks of real fractures and creating sinusoidal shaped fractures. The trained model is evaluated on real core images from two boreholes and provides an average precision of approximately 95%. The identified fractures are further analyzed and compared to manually segmented fractures in terms of fracture dip angle and dip direction, which achieved average absolute errors of around 2° and 11°, respectively. Overall, the study presents a novel application of an advanced machine learning algorithm for fracture detection and analysis from unwrapped core images.

Diagnostic Accuracy of 3D Ultrasound and Artificial Intelligence for Detection of Pediatric Wrist Injuries.

Wrist trauma is common in children, typically requiring radiography for diagnosis and treatment planning. However, many children do not have fractures and are unnecessarily exposed to radiation. Ultrasound performed at bedside could detect fractures prior to radiography. Modern tools including three-dimensional ultrasound (3DUS) and artificial intelligence (AI) have not yet been applied to this task. Our purpose was to assess (1) feasibility, reliability, and accuracy of 3DUS for detection of pediatric wrist fractures, and (2) accuracy of automated fracture detection via AI from 3DUS sweeps. Children presenting to an emergency department with unilateral upper extremity injury to the wrist region were scanned on both the affected and unaffected limb. Radiographs of the symptomatic limb were obtained for comparison. Ultrasound scans were read by three individuals to determine reliability. An AI network was trained and compared against the human readers. Thirty participants were enrolled, resulting in scans from fifty-five wrists. Readers had a combined sensitivity of 1.00 and specificity of 0.90 for fractures. AI interpretation was indistinguishable from human interpretation, with all fractures detected in the test set of 36 images (sensitivity = 1.0). The high sensitivity of 3D ultrasound and automated AI ultrasound interpretation suggests that ultrasound could potentially rule out fractures in the emergency department.

Automatic Detection of Fractures Based on Optimal Path Search in Well Logging Images

Reservoir fractures are essential locations to gather oil and gas. Recently, imaging logging technology has become a mainstream method for obtaining stratigraphic information. This paper proposed a combined optimal path search strategy to effectively identify and extract the fracture information in well logging images. Specifically, the threshold segmentation of logging images is used to obtain the binary image. In addition, the identification of connected fractures in the logging image is transformed into the optimal path search, and the identification and extraction of reservoir fractures are realized by constructing the optimal path between the two ends of fractures. Finally, an improved ant colony algorithm is applied to filter irrelevant information and extract fractures automatically by recording all the ants’ exploration trajectories in the ant colony. Compared with previous approaches, the proposed method can eliminate irrelevant background features and merely reserve pixels corresponding to fractures. Simultaneously, relative to the conventional strategy, the time consumption is reduced by more than 98%. The findings of this study can help for better extracting fractures automatically and reducing manual workload.

Bone fracture detection through the two-stage system of Crack-Sensitive Convolutional Neural Network

Abstract Automated fracture detection is an essential part in a computer-aided tele-medicine system. Fractures often occur in human's arbitrary bone due to accidental injuries such as slipping. In fact, many hospitals lack experienced surgeons to diagnose fractures. Therefore, computer-aided diagnosis (CAD) reduces the burden on doctors and identifies fracture. We present a new classification network, Crack-Sensitive Convolutional Neural Network (CrackNet), which is sensitive to fracture lines. In this paper, we propose a new two-stage system to detect fracture. Firstly, we use Faster Region with Convolutional Neutral Network (Faster R-CNN) to detect 20 different types of bone regions in X-ray images, and then we recognize whether each bone region is fractured by using CrackNet. Total of 1052 images are used to test our system, of which 526 are fractured images and the rest are non-fractured images. We assess the performance of our proposed system with X-ray images from Haikou People's Hospital, achieving 90.11% accuracy and 90.14% F-measure. And our system is better than other two-stage systems.

Mapping the fracture network in the Lilstock pavement, Bristol Channel, UK: manual versus automatic

Abstract. The 100 000 m2 wave-cut pavement in the Bristol Channel near Lilstock, UK, is a world-class outcrop, perfectly exposing a very large fracture network in several thin limestone layers. We present an analysis based on manual interpretation of fracture generations in selected domains and compare it with automated fracture tracing. Our dataset of high-resolution aerial photographs of the complete outcrop was acquired by an unmanned aerial vehicle, using a survey altitude optimized to resolve all fractures. We map fractures and identify fracture generations based on abutting and overprinting criteria, and we present the fracture networks of five selected representative domains. Each domain is also mapped automatically using ridge detection based on the complex shearlet transform method. The automatic fracture detection technique provides results close to the manually traced fracture networks in shorter time but with a bias towards closely spaced Y over X nodes. The assignment of fractures into generations cannot yet be done automatically, because the fracture traces extracted by the automatic method are segmented at the nodes, unlike the manual interpretation in which fractures are traced as a path from fracture tip to fracture tip and consist of several connected segments. This segmentation makes an interpretation of relative age impossible, because the identification of correct abutting relationships requires the investigation of the complete fracture trace by following a clearly defined set of rules. Generations 1 and 2 are long fractures that traverse all domains. Generation 3 is only present in the southwestern domains. Generation 4 follows an ENE–WSW striking trend, is suborthogonal to generations 1 and 2, and abuts on them and generation 3, if present. Generations 5 is the youngest fracture set with a range of orientations, creating polygonal patterns by abutting at all other fracture generations. Our mapping results show that the northeastern domains only contain four fracture generations; thus, the five generations of the outcrop identified in the southwestern domains are either not all present in each of the five domains or vary locally in their geometry, preventing the interpreter from linking the fractures to their respective generation over several spatially separate mapping domains. Fracture intensities differ between domains where the lowest is in the NE with 7.3 m−1 and the highest is in the SW with 10 m−1, coinciding with different fracture orientations and distributions of abutting relationships. Each domain has slightly different fracture network characteristics, and greater connectivity occurs where the development of later shorter fractures is not affected by the stress shadowing of pre-existing longer fractures.

Is Deep Learning On Par with Human Observers for Detection of Radiographically Visible and Occult Fractures of the Scaphoid?

Preliminary experience suggests that deep learning algorithms are nearly as good as humans in detecting common, displaced, and relatively obvious fractures (such as, distal radius or hip fractures). However, it is not known whether this also is true for subtle or relatively nondisplaced fractures that are often difficult to see on radiographs, such as scaphoid fractures. (1) What is the diagnostic accuracy, sensitivity, and specificity of a deep learning algorithm in detecting radiographically visible and occult scaphoid fractures using four radiographic imaging views? (2) Does adding patient demographic (age and sex) information improve the diagnostic performance of the deep learning algorithm? (3) Are orthopaedic surgeons better at diagnostic accuracy, sensitivity, and specificity compared with deep learning? (4) What is the interobserver reliability among five human observers and between human consensus and deep learning algorithm? We retrospectively searched the picture archiving and communication system (PACS) to identify 300 patients with a radiographic scaphoid series, until we had 150 fractures (127 visible on radiographs and 23 only visible on MRI) and 150 non-fractures with a corresponding CT or MRI as the reference standard for fracture diagnosis. At our institution, MRIs are usually ordered for patients with scaphoid tenderness and normal radiographs, and a CT with radiographically visible scaphoid fracture. We used a deep learning algorithm (a convolutional neural network [CNN]) for automated fracture detection on radiographs. Deep learning, an advanced subset of artificial intelligence, combines artificial neuronal layers to resemble a neuron cell. CNNs-essentially deep learning algorithms resembling interconnected neurons in the human brain-are most commonly used for image analysis. Area under the receiver operating characteristic curve (AUC) was used to evaluate the algorithm's diagnostic performance. An AUC of 1.0 would indicate perfect prediction, whereas 0.5 would indicate that a prediction is no better than a flip of a coin. The probability of a scaphoid fracture generated by the CNN, sex, and age were included in a multivariable logistic regression to determine whether this would improve the algorithm's diagnostic performance. Diagnostic performance characteristics (accuracy, sensitivity, and specificity) and reliability (kappa statistic) were calculated for the CNN and for the five orthopaedic surgeon observers in our study. The algorithm had an AUC of 0.77 (95% CI 0.66 to 0.85), 72% accuracy (95% CI 60% to 84%), 84% sensitivity (95% CI 0.74 to 0.94), and 60% specificity (95% CI 0.46 to 0.74). Adding age and sex did not improve diagnostic performance (AUC 0.81 [95% CI 0.73 to 0.89]). Orthopaedic surgeons had better specificity (0.93 [95% CI 0.93 to 0.99]; p < 0.01), while accuracy (84% [95% CI 81% to 88%]) and sensitivity (0.76 [95% CI 0.70 to 0.82]; p = 0.29) did not differ between the algorithm and human observers. Although the CNN was less specific in diagnosing relatively obvious fractures, it detected five of six occult scaphoid fractures that were missed by all human observers. The interobserver reliability among the five surgeons was substantial (Fleiss' kappa = 0.74 [95% CI 0.66 to 0.83]), but the reliability between the algorithm and human observers was only fair (Cohen's kappa = 0.34 [95% CI 0.17 to 0.50]). Initial experience with our deep learning algorithm suggests that it has trouble identifying scaphoid fractures that are obvious to human observers. Thirteen false positive suggestions were made by the CNN, which were correctly detected by the five surgeons. Research with larger datasets-preferably also including information from physical examination-or further algorithm refinement is merited. Level III, diagnostic study.

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

To demonstrate the feasibility and performance of an object detection convolutional neural network (CNN) for fracture detection and localization on wrist radiographs. 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. 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. 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.

Evaluation of a workflow to derive terrestrial light detection and ranging fracture statistics of a tight gas sandstone reservoir analog

Understanding natural fracture networks in the subsurface is highly challenging, as direct one-dimensional borehole data are unable to reflect their spatial complexity, and three-dimensional seismic data are limited in spatial resolution to resolve individual meter-scale fractures. Here, we present a prototype workflow for automated fracture detection along horizontal scan lines using terrestrial light detection and ranging (t-LIDAR). Data are derived from a kilometer-scale Pennsylvanian (locally upper Carboniferous) reservoir outcrop analog in the Lower Saxony Basin, northwestern Germany. The workflow allows the t-LIDAR data to be integrated into conventional reservoir-modeling software for characterizing natural fracture networks with regard to orientation and spatial distribution. The analysis outlines the lateral reorientation of fractures from a west–southwest/east–northeast strike, near a normal fault with approximately 600 m (∼1970 ft) displacement, toward an east–west strike away from the fault. Fracture corridors, 10–20 m (33–66 ft) wide, are present in unfaulted rocks with an average fracture density of 3.4–3.9 m−1 (11.2–12.8 ft−1). A reservoir-scale digital outcrop model was constructed as a basis for data integration. The fracture detection and analysis serve as input for a stochastically modeled discrete fracture network, demonstrating the transferability of the derived data into standard hydrocarbon exploration-and-production-industry approaches. The presented t-LIDAR workflow provides a powerful tool for quantitative spatial analysis of outcrop analogs, in terms of natural fracture network characterization, and enriches classical outcrop investigation techniques. This study may contribute to a better application of outcrop analog data to naturally fractured reservoirs in the subsurface, reducing uncertainties in the characterization of this reservoir type at depth.

Machine-learning-based object detection in images for reservoir characterization: A case study of fracture detection in shales

Imaging tools are widely used in the petroleum industry to investigate structural features of reservoir rocks directly at multiple scales. Quantitative image analysis is often used to determine various rock properties, but it requires significant time and effort, particularly to analyze a large number of samples. Automated object detection represents a potential solution to this efficiency problem. This method uses computers to efficiently provide quantitative information for thousands of images. Automated fracture detection in scanning electron microscope (SEM) images is presented as an example to show the workflow of using advanced deep-learning tools for quantitative rock characterization. First, an automatic object-detection method is presented for fast identification and characterization of microfractures in shales. Using this approach, we analyzed 100 SEM images obtained from deformed and intact samples of a carbonate-rich shale and a siliceous shale with the goal of analyzing the abundance and characteristics of microfractures generated during hydraulic fracturing. Most of the fractures are detected with about 90% success rate relative to manual picking. Second, we obtained statistics of length and areal porosities of these fractures. The experimentally deformed samples had slightly more detectable microfractures (1.8 fractures/image on average compared to 1.6 fractures/image), and the microfractures induced by shear deformation tend to be short (&lt;50 μm) in the Eagle Ford and long in the siliceous samples, presumably because of differences in rock fabric. In future work, this approach will be applied to characterize the shape and size of mineral grains and to analyze relationships between fractures and minerals.

Tectonic stress estimation from ultrasonic borehole image logs in a coal bed methane well, northeastern Qinshui Basin, China

Small-scale structures and tectonic stress are known to affect the production of coalbed methane. Ultrasonic borehole image logging (UBI) can be used to identify fractures and breakout zones around the wellbore and to analyze tectonic stress fields. In this study, a variety of image processing and automated fracture detection methods were used to process and interpret the ultrasonic logging data in a coal bed methane well. Combined with other logging data in the well, this study analyzed the ultrasonic logging response characteristics of the formation lithology and structure. Statistical analysis has been completed to identify the fracturing pattern. The results showed that the dip angle and orientation of the natural fracture were related to regional tectonics. The paleo-stress field, reflected by the natural fractures, was consistent with the geological conditions, while the current stress field, reflected by the breakout zones, was found to be different from the paleo-stress field. This study proposed a new method by using fractures with associated conjugate joints, which are obtained from UBI logs, to estimate the paleo-stress orientation.