Traumatic fractures in adults: missed diagnosis on plain radiographs in the Emergency Department

Radiographic Detection of Post-Traumatic Bone Fractures: Contribution of Artificial Intelligence Software to the Analysis of Senior and Junior Radiologists

Objectives

The aims of this study were: (a) to evaluate the performance of an artificial intelligence (AI) software package (Boneview Trauma, Gleamer) for the detection of post-traumatic bone fractures in radiography as a standalone; (b) used by two radiologists (osteoarticular senior and junior); and (c) to determine to whom AI would be most helpful.

Materials and Methods

Within 14 days of a trauma, 101 consecutive patients underwent radiographic examination of the upper or lower limbs. The definite diagnosis for identifying fractures was: (a) radio-clinical consensus between the radiologist on-call who analyzed the images and the orthopedist (Group 1); (b) Cone Beam computed tomography (CBCT) exploration of the area of interest, in case of doubts or absence of consensus (Group 2). Independently of this diagnosis for both groups, the radiographic images were separately analyzed by two radiologists (osteoarticular senior: SR; junior: JR) prior without, and thereafter with the results of AI.

Results

AI performed better than the radiologists in detecting common fractures (Group 1), but not subtle fractures (Group 2). In association with AI, both radiologists increased their overall performances in both groups, whereas this increase was significantly higher for the JR (p < 0.05).

Conclusion

AI is reliable for common radiographic fracture identification and is a useful learning tool for radiologists in training. However, the software's overall performance does not exceed that of an osteoarticular senior radiologist, particularly in case of subtle lesions.

Artificial intelligence diagnostic model for multi-site fracture X-ray images of extremities based on deep convolutional neural networks

Background

The rapid and accurate diagnosis of fractures is crucial for timely treatment of trauma patients. Deep learning, one of the most widely used forms of artificial intelligence (AI), is now commonly employed in medical imaging for fracture detection. This study aimed to construct a deep learning model using big data to recognize multiple-fracture X-ray images of extremity bones.

Methods

Radiographic imaging data of extremities were retrospectively collected from five hospitals between January 2017 and September 2020. The total number of people finally included was 25,635 and the total number of images included was 26,098. After labeling the lesions, the randomized method used 90% of the data as the training set to develop the fracture detection model, and the remaining 10% was used as the validation set to verify the model. The faster region convolutional neural networks (R-CNN) algorithm was adopted to construct diagnostic models for detection. The Dice coefficient was used to evaluate the image segmentation accuracy. The performances of detection models were evaluated with sensitivity, specificity, and area under the receiver operating characteristic curve (AUC).

Results

The free-response receiver operating characteristic (FROC) curve value was 0.886 and 0.843 for the detection of single and multiple fractures, respectively. Additionally, the effective identification AUC for all parts was higher than 0.920. Notably, the AUC for wrist fractures reached 0.952. The average accuracy in detecting bone fracture regions in the extremities was 0.865. When analyzing single and multiple lesions at the patient level, the sensitivity was 0.957 for patients with multiple lesions and 0.852 for those with single lesions. In the segmentation task, the training set (the data set used by the machine learning model to train and learn) and the validation set (the data set used to evaluate the performance of the model) reached 0.996 and 0.975, respectively.

Conclusions

The faster R-CNN training algorithm exhibits excellent performance in simultaneously identifying fractures in the hands, feet, wrists, ankles, radius and ulna, and tibia and fibula on X-ray images. It demonstrates high accuracy, low false-negative rates, and controllable false-positive rates. It can serve as a valuable screening tool.

Cinematic rendering improves the AO/OTA classification of distal femur fractures compared to volume rendering: a retrospective single-center study

Purpose: Correctly classifying distal femur fractures is essential for surgical treatment planning and patient prognosis. This study assesses the potential of Cinematic Rendering (CR) in classifying these fractures, emphasizing its reported ability to produce more realistic images than Volume Rendering (VR). Methods: Data from 88 consecutive patients with distal femoral fractures collected between July 2013 and July 2020 were included. Two orthopedic surgeons independently evaluated the fractures using CR and VR. The inter-rater and intra-rater agreement was evaluated by using the Cicchetti-Allison weighted Kappa method. Accuracy, precision, recall, and F1 score were also calculated. Diagnostic confidence scores (DCSs) for both imaging methods were compared using chi-square or Fisher's exact tests. Results: CR reconstruction yielded excellent inter-observer (Kappa = 0.989) and intra-observer (Kappa = 0.992) agreement, outperforming VR (Kappa = 0.941 and 0.905, respectively). While metrics like accuracy, precision, recall, and F1 scores were higher for CR, the difference was not statistically significant (p > 0.05). However, DCAs significantly favored CR (p < 0.05). Conclusion: CR offers a superior visualization of distal femur fractures than VR. It enhances fracture classification accuracy and bolsters diagnostic confidence. The high inter- and intra-observer agreement underscores its reliability, suggesting its potential clinical importance.

A review on artificial intelligence for the diagnosis of fractures in facial trauma imaging

Patients with facial trauma may suffer from injuries such as broken bones, bleeding, swelling, bruising, lacerations, burns, and deformity in the face. Common causes of facial-bone fractures are the results of road accidents, violence, and sports injuries. Surgery is needed if the trauma patient would be deprived of normal functioning or subject to facial deformity based on findings from radiology. Although the image reading by radiologists is useful for evaluating suspected facial fractures, there are certain challenges in human-based diagnostics. Artificial intelligence (AI) is making a quantum leap in radiology, producing significant improvements of reports and workflows. Here, an updated literature review is presented on the impact of AI in facial trauma with a special reference to fracture detection in radiology. The purpose is to gain insights into the current development and demand for future research in facial trauma. This review also discusses limitations to be overcome and current important issues for investigation in order to make AI applications to the trauma more effective and realistic in practical settings. The publications selected for review were based on their clinical significance, journal metrics, and journal indexing.

Influence of residency discipline and seniority on traumatic musculoskeletal radiographs interpretation accuracy: a multicenter study

OBJECTIVES

Imaging studies are a significant and integral part of the initial assessment of patients admitted to the emergency department. Developing imaging diagnostic abilities early in residency is of paramount importance. The purpose of this study was to evaluate and compare diagnosis accuracy of common musculoskeletal X-rays (XR) between residency disciplines and seniority.

METHODS

A multicenter study which evaluated orthopedic surgery, emergency medicine (EM), and radiology residents, through a test set of common MSK XR. Residents were classified as "beginner" or "advanced" according to postgraduate year per residency. Residents were asked to answer whether the radiograph shows normal or pathological findings (success rate) and what is the diagnosis ("diagnosis accuracy"). Residents' answers were analyzed and assessed compared to experts' consensus.

RESULTS

A total of 100 residents (62% beginners) participated in this study. Fifty-four were orthopedic surgeons, 29 were EM residents and 17 were radiologists. The entire cohort's overall success rate was 88.5%. The overall mean success rates for orthopedic, EM, and radiology residents were 93.2%, 82.8%, and 83.3%, respectively, and were significantly different (p < 0.0001). Orthopedic residents had significantly higher diagnostic accuracy rates compared with both radiology and EM residents (p < 0.001). Advanced orthopedic and EM residents demonstrated higher diagnostic accuracy rates compared to beginner residents (p = 0.001 and p = 0.03, respectively).

CONCLUSION

Orthopedic residents presented higher diagnosis accuracy of MSK imaging compared to EM and radiology residents. Seniority had a positive effect on diagnosis accuracy. The development of an educational program on MSK XR is necessary to enhance the competency of physicians in their daily practice.

Robust Multi-View Fracture Detection in the Presence of Other Abnormalities Using HAMIL-Net

INTRODUCTION

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.

MATERIALS AND METHODS

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.

RESULTS

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.

CONCLUSIONS

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.

Commercially-available AI algorithm improves radiologists' sensitivity for wrist and hand fracture detection on X-ray, compared to a CT-based ground truth

OBJECTIVES

Algorithms for fracture detection are spreading in clinical practice, but the use of X-ray-only ground truth can induce bias in their evaluation. This study assessed radiologists' performances to detect wrist and hand fractures on radiographs, using a commercially-available algorithm, compared to a computerized tomography (CT) ground truth.

METHODS

Post-traumatic hand and wrist CT and concomitant X-ray examinations were retrospectively gathered. Radiographs were labeled based on CT findings. The dataset was composed of 296 consecutive cases: 118 normal (39.9%), 178 pathological (60.1%) with a total of 267 fractures visible in CT. Twenty-three radiologists with various levels of experience reviewed all radiographs without AI, then using it, blinded towards CT results.

RESULTS

Using AI improved radiologists' sensitivity (Se, 0.658 to 0.703, p < 0.0001) and negative predictive value (NPV, 0.585 to 0.618, p < 0.0001), without affecting their specificity (Sp, 0.885 vs 0.891, p = 0.91) or positive predictive value (PPV, 0.887 vs 0.899, p = 0.08). On the radiographic dataset, based on the CT ground truth, stand-alone AI performances were 0.771 (Se), 0.898 (Sp), 0.684 (NPV), 0.915 (PPV), and 0.764 (AUROC) which were lower than previously reported, suggesting a potential underestimation of the number of missed fractures in the AI literature.

CONCLUSIONS

AI enabled radiologists to improve their sensitivity and negative predictive value for wrist and hand fracture detection on radiographs, without affecting their specificity or positive predictive value, compared to a CT-based ground truth. Using CT as gold standard for X-ray labels is innovative, leading to algorithm performance poorer than reported elsewhere, but probably closer to clinical reality.

CLINICAL RELEVANCE STATEMENT

Using an AI algorithm significantly improved radiologists' sensitivity and negative predictive value in detecting wrist and hand fractures on radiographs, with ground truth labels based on CT findings.

KEY POINTS

• Using CT as a ground truth for labeling X-rays is new in AI literature, and led to algorithm performance significantly poorer than reported elsewhere (AUROC: 0.764), but probably closer to clinical reality. • AI enabled radiologists to significantly improve their sensitivity (+ 4.5%) and negative predictive value (+ 3.3%) for the detection of wrist and hand fractures on X-rays. • There was no significant change in terms of specificity or positive predictive value.

Lisfranc Sports Injuries: What Do We Know So Far?

Lisfranc sports injuries include tarsometatarsal joint injuries, which may be accompanied by fractures. They most commonly occur due to a blow or axial force. The aim of this review is to assess the current standards for surgical intervention in Lisfranc injuries resulting from sports-related accidents. This evaluation will cover the timing of treatment, the recovery process, and the appropriate timing for a return to normal sporting activities. This research was done via an analytical review of current literature. Methods included a structured search strategy on PubMed, Science Direct, and Google Scholar. The collated literature was processed using formal inclusion or exclusion, data extraction, and validity assessment. Joint involvement and severity were taken into account while classifying Lisfranc injuries. The primary fixation and fusion techniques for Lisfranc injuries were compared, and the surgical management of these injuries was examined in all of the literature. Treatment recovery times were examined, and the results were talked about. A variety of injuries, from minor sprains to serious fractures and rips, make up Lisfranc injuries. Although open reduction internal fixation (ORIF) in combination with primary arthrodesis (PA) is now thought to be the optimum course of treatment, its acceptance has increased. Patients with Lisfranc injuries can usually expect excellent outcomes and the return of joint function to its pre-injury form if the injury is appropriately assessed and treated. Lisfranc injuries are manageable and have a good recovery time if not neglected. The outcomes of management and surgical options are also quite satisfactory.

Application of Artificial Intelligence and Machine Learning in Diagnosing Scaphoid Fractures: A Systematic Review

The integration of artificial intelligence (AI) in healthcare has sparked interest in its potential to revolutionize medical diagnostics. This systematic review explores the application of AI and machine learning (ML) techniques in diagnosing scaphoid fractures, which account for a significant percentage of carpal bone fractures and have important implications for wrist function. Scaphoid fractures, common in young and active individuals, require an early and accurate diagnosis for effective treatment. AI has the potential to automate and improve the accuracy of scaphoid fracture detection on radiography, aiding in early diagnosis and reducing unnecessary clinical examinations. This systematic review discusses the methods used to identify relevant studies, including search criteria and quality assessment tools, and presents the results of the selected studies. The findings indicate that AI-driven methods can improve diagnostic accuracy, reducing the risk of missed fractures and complications. AI assistance can also alleviate the workload of medical professionals, improving diagnostic efficiency and reducing observer fatigue. However, challenges such as algorithm limitations and the need for continuous refinement must be addressed to ensure reliable fracture identification. This review underscores the clinical significance of AI-assisted diagnostics, especially in cases where fractures may be subtle or occult. It emphasizes the importance of integrating AI into medical education and training and calls for robust data collection and collaboration between AI developers and medical practitioners. Future research should focus on larger datasets, algorithm improvement, cost-effectiveness assessment, and international partnerships to fully harness the potential of AI in diagnosing scaphoid fractures.

AC-Faster R-CNN: an improved detection architecture with high precision and sensitivity for abnormality in spine x-ray images

Objective.In clinical medicine, localization and identification of disease on spinal radiographs are difficult and require a high level of expertise in the radiological discipline and extensive clinical experience. The model based on deep learning acquires certain disease recognition abilities through continuous training, thereby assisting clinical physicians in disease diagnosis. This study aims to develop an object detection network that accurately locates and classifies the abnormal parts in spinal x-ray photographs.Approach.This study proposes a deep learning-based automated multi-disease detection architecture called Abnormality Capture-Faster Region-based Convolutional Neural Network (AC-Faster R-CNN), which develops the feature fusion structure Deformable Convolution Feature Pyramid Network and the abnormality capture structure Abnormality Capture Head. Through the combination of dilated and deformable convolutions, the model better captures the multi-scale information of lesions. To further improve the detection performance, the contrast enhancement algorithm Contrast Limited Adaptive Histogram Equalization is used for image preprocessing.Main results.The proposed model is extensively evaluated on a testing set containing 1007 spine x-ray images and the experimental results show that the AC-Faster R-CNN architecture outperforms the baseline model and other advanced detection architectures. The mean Average Precision at Intersection over Union of 50% are 39.8%, the Precision and Sensitivity at the optimal cutoff point of Precision-Recall curve are 48.6% and 46.3%, respectively, reaching the current state-of-the-art detection level.Significance.AC-Faster R-CNN exhibits high precision and sensitivity in abnormality detection tasks of spinal x-ray images, and effectively locates and identifies abnormal areas. Additionally, this study would provide reference and comparison for the further development of medical automatic detection.

Diagnostic Accuracy and Confidence in Management of Forearm and Hand Fractures Among Foundation Doctors in the Accident and Emergency Department: Survey Study

BACKGROUND

Accurate interpretation of radiographs is crucial for junior doctors in the accident and emergency (A&E) department (the emergency medicine department). However, it remains a significant challenge and a leading cause of diagnostic errors.

OBJECTIVE

This study aimed to evaluate the accuracy and confidence of foundation doctors (doctors within their first 2 years of qualifying) in correctly interpreting and managing forearm and hand fractures on plain radiographs.

METHODS

A total of 42 foundation doctors with less than 2 years of experience and no prior emergency medicine training who worked in a large district general hospital participated in a web-based questionnaire. The questionnaire consisted of 3 case studies: distal radius fracture, scaphoid fracture, and a normal radiograph. Respondents were required to identify the presence or absence of a fracture, determine the fracture location, suggest appropriate management, and rate their confidence on a Likert scale.

RESULTS

Overall, 48% (61/126) of respondents accurately identified the presence and location of fractures. The correct management option was chosen by 64% (81/126) of respondents. The median diagnostic confidence score was 4 of 10, with a mean diagnostic certainty of 4.4 of 10. Notably, respondents exhibited a significantly lower confidence score for the normal radiograph compared to the distal radius fracture radiograph (P=.01).

CONCLUSIONS

This study reveals diagnostic uncertainty among foundation doctors in interpreting plain radiographs, with a notable inclination toward overdiagnosing fractures. The findings emphasize the need for close supervision and senior support to mitigate diagnostic errors. Further training and educational interventions are warranted to improve the accuracy and confidence of junior doctors in radiographic interpretation. This study has several limitations, including a small sample size and reliance on self-reported data. The findings may not be generalizable to other health care settings or specialties. Future research should aim for larger, more diverse samples and explore the impact of specific educational interventions on diagnostic accuracy and confidence.

Deep convolutional neural network for rib fracture recognition on chest radiographs

Introduction

Rib fractures are a prevalent injury among trauma patients, and accurate and timely diagnosis is crucial to mitigate associated risks. Unfortunately, missed rib fractures are common, leading to heightened morbidity and mortality rates. While more sensitive imaging modalities exist, their practicality is limited due to cost and radiation exposure. Point of care ultrasound offers an alternative but has drawbacks in terms of procedural time and operator expertise. Therefore, this study aims to explore the potential of deep convolutional neural networks (DCNNs) in identifying rib fractures on chest radiographs.

Methods

We assembled a comprehensive retrospective dataset of chest radiographs with formal image reports documenting rib fractures from a single medical center over the last five years. The DCNN models were trained using 2000 region-of-interest (ROI) slices for each category, which included fractured ribs, non-fractured ribs, and background regions. To optimize training of the deep learning models (DLMs), the images were segmented into pixel dimensions of 128 × 128.

Results

The trained DCNN models demonstrated remarkable validation accuracies. Specifically, AlexNet achieved 92.6%, GoogLeNet achieved 92.2%, EfficientNetb3 achieved 92.3%, DenseNet201 achieved 92.4%, and MobileNetV2 achieved 91.2%.

Discussion

By integrating DCNN models capable of rib fracture recognition into clinical decision support systems, the incidence of missed rib fracture diagnoses can be significantly reduced, resulting in tangible decreases in morbidity and mortality rates among trauma patients. This innovative approach holds the potential to revolutionize the diagnosis and treatment of chest trauma, ultimately leading to improved clinical outcomes for individuals affected by these injuries. The utilization of DCNNs in rib fracture detection on chest radiographs addresses the limitations of other imaging modalities, offering a promising and practical solution to improve patient care and management.

Automatic classification of distal radius fracture using a two-stage ensemble deep learning framework

Distal radius fractures (DRFs) are one of the most common types of wrist fracture and can be subdivided into intra- and extra-articular fractures. Compared with extra-articular DRFs which spare the joint surface, intra-articular DRFs extend to the articular surface and can be more difficult to treat. Identification of articular involvement can provide valuable information about the characteristics of fracture patterns. In this study, a two-stage ensemble deep learning framework was proposed to differentiate intra- and extra-articular DRFs automatically on posteroanterior (PA) view wrist X-rays. The framework firstly detects the distal radius region of interest (ROI) using an ensemble model of YOLOv5 networks, which imitates the clinicians' search pattern of zooming in on relevant regions to assess abnormalities. Secondly, an ensemble model of EfficientNet-B3 networks classifies the fractures in the detected ROIs into intra- and extra-articular. The framework achieved an area under the receiver operating characteristic curve of 0.82, an accuracy of 0.81, a true positive rate of 0.83 and a false positive rate of 0.27 (specificity of 0.73) for differentiating intra- from extra-articular DRFs. This study has demonstrated the potential in automatic DRF characterization using deep learning on clinically acquired wrist radiographs and can serve as a baseline for further research in incorporating multi-view information for fracture classification.

Deep Learning in the Detection of Rare Fractures - Development of a "Deep Learning Convolutional Network" Model for Detecting Acetabular Fractures

BACKGROUND

Fracture detection by artificial intelligence and especially Deep Convolutional Neural Networks (DCNN) is a topic of growing interest in current orthopaedic and radiological research. As learning a DCNN usually needs a large amount of training data, mostly frequent fractures as well as conventional X-ray are used. Therefore, less common fractures like acetabular fractures (AF) are underrepresented in the literature. The aim of this pilot study was to establish a DCNN for detection of AF using computer tomography (CT) scans.

METHODS

Patients with an acetabular fracture were identified from the monocentric consecutive pelvic injury registry at the BG Trauma Center XXX from 01/2003 - 12/2019. All patients with unilateral AF and CT scans available in DICOM-format were included for further processing. All datasets were automatically anonymised and digitally post-processed. Extraction of the relevant region of interests was performed and the technique of data augmentation (DA) was implemented to artificially increase the number of training samples. A DCNN based on Med3D was used for autonomous fracture detection, using global average pooling (GAP) to reduce overfitting.

RESULTS

From a total of 2,340 patients with a pelvic fracture, 654 patients suffered from an AF. After screening and post-processing of the datasets, a total of 159 datasets were enrolled for training of the algorithm. A random assignment into training datasets (80%) and test datasets (20%) was performed. The technique of bone area extraction, DA and GAP increased the accuracy of fracture detection from 58.8% (native DCNN) up to an accuracy of 82.8% despite the low number of datasets.

CONCLUSION

The accuracy of fracture detection of our trained DCNN is comparable to published values despite the low number of training datasets. The techniques of bone extraction, DA and GAP are useful for increasing the detection rates of rare fractures by a DCNN. Based on the used DCNN in combination with the described techniques from this pilot study, the possibility of an automatic fracture classification of AF is under investigation in a multicentre study.

The FRISK (Fracture Risk)-A New Tool to Indicate the Probability of Fractures

Increasing patient inflow into the emergency department makes it necessary to optimize triage management. The scope of this work was to determine simple factors that could detect fractures in patients without the need for specialized personnel. Between 2014 and 2015, 798 patients were admitted to an orthopedic emergency department and prospectively included in the study. The patients received a questionnaire before contacting the doctor. Objective and subjective data were evaluated to determine fracture risk for the upper and lower extremities. The highest risk for fractures in one region was the hip (73.21%; n = 56), followed by the wrist (60.32%; n = 63) and the femoral shaft (4 of 7, 57.14%; n = 7). The regions with the lowest risk were the knee (8.41%; n = 107), the ankle (18.29%; n = 164), and the forearm shaft (30.00%; n = 10). Age was a predictor for fracture: patients older than 59 years had a risk greater than 59.26%, and patients older than 90 years had a risk greater than 83.33%. The functional questions could exclude fractures. Three factors seem to be able to predict fracture risk: the injured region, the patient's age, and a functional question. They can be used for a probatory heuristic that needs to be proven in a prospective way.

Mapping of Orthopaedic Fractures for Optimal Surgical Guidance

Orthopaedic fractures may be difficult to treat surgically if accurate information on the fracture propagation and its exit points are not known. Even with two-dimensional (2D) radiographic images, it is difficult to be completely certain of the exact location of the fracture site, the fracture propagation pattern and the exit points of the fracture. Three-dimensional (3D) computerised tomographic models are better in providing surgeons with the extent of bone fractures, but they may still not be sufficient to allow surgeons to plan open reduction and internal fixation (ORIF) surgery.Fracture patterns and fracture maps are developed to be visual tools in 2D and 3D. These tools can be developed using fractured bones either before or after fracture reduction. Aside from being beneficial to surgeons during pre-surgical planning, these maps aid bioengineers who design fracture fixation plates and implants for these fractures, as well as represent fracture classifications.Fracture maps can be either created ex silico or in silico. Ex silico models are created using 3D printed bone models, onto which fracture patterns are marked. In silico fracture models are created by tracing the fracture lines from a fractured bone to a healthy bone template on a computer. The points of interest in both of these representations are the path of fracture propagation on the bone's surface and exit zones, which eventually determine the surgeon's choice of plate and fracture reduction. Both ex silico and in silico fracture maps are used for pre-surgical planning by the surgeons where they can plan the best way to reduce the fracture as well as template various implants in a low-risk environment before performing the surgery.Recently, fracture maps have been further digitised into heat maps. These heat maps provide visual representations of critical regions of fractures propagating through the bone and identify the weaker zones in the bone structure. These heat maps can allow engineers to develop optimal surgical plates to fix an array of fracture patterns propagating through the bone. Correlation of fractured regions with the mechanisms of injury, age, gender, etc. may improve fracture predictability in the future and optimise the intervention, along with making sure that surgeons do not miss fractures of the bone that may otherwise be hidden from plain sight.

Effect of Vacuum Sealing Drainage on Soft Tissue Injury of Traumatic Fracture and Its Effect on Wound Recovery

Purpose

The current work is mainly to explore the effect of vacuum sealing drainage (VSD) on soft tissue injury (STI) caused by traumatic fractures (TFs) and its effect on wound recovery.

Methods

We first selected 90 patients with TF STI from May 2019 to May 2021, of which 40 patients (control group) received routine treatment, and the other 50 patients (observation group) were treated with VSD. The curative effect, rehabilitation (changing dressing frequency, healing time, and hospitalization time), pain severity, patient comfort, and complications were evaluated and compared.

Results

The observation group exhibited a higher total effective rate, lower dressing change frequency, complication rate, and shorter healing time and hospital stay than the control group, which are statistically significant. Statistically milder pain sensation and better patient comfort were also determined in the observation group.

Conclusions

VSD is effective and safe in the treatment of TF-induced sexually transmitted infections, which can effectively accelerate wound recovery while reducing pain sensation and improving patient comfort, with clinical promotion value.

Detection of acute rib fractures on CT images with convolutional neural networks: effect of location and type of fracture and reader's experience

PURPOSE

The evaluation of all ribs on thin-slice CT images is time consuming and it can be difficult to accurately assess the location and type of rib fracture in an emergency. The aim of our study was to develop and validate a convolutional neural network (CNN) algorithm for the detection of acute rib fractures on thoracic CT images and to investigate the effect of the CNN algorithm on radiologists' performance.

METHODS

The dataset for development of a CNN consisted of 539 thoracic CT scans with 4906 acute rib fractures. A three-dimensional faster region-based CNN was trained and evaluated by using tenfold cross-validation. For an observer performance study to investigate the effect of CNN outputs on radiologists' performance, 30 thoracic CT scans (28 scans with 90 acute rib fractures and 2 without rib fractures) which were not included in the development dataset were used. Observer performance study involved eight radiologists who evaluated CT images first without and second with CNN outputs. The diagnostic performance was assessed by using figure of merit (FOM) values obtained from the jackknife free-response receiver operating characteristic (JAFROC) analysis.

RESULTS

When radiologists used the CNN output for detection of rib fractures, the mean FOM value significantly increased for all readers (0.759 to 0.819, P = 0.0004) and for displaced (0.925 to 0.995, P = 0.0028) and non-displaced fractures (0.678 to 0.732, P = 0.0116). At all rib levels except for the 1st and 12th ribs, the radiologists' true-positive fraction of the detection became significantly increased by using the CNN outputs.

CONCLUSION

The CNN specialized for the detection of acute rib fractures on CT images can improve the radiologists' diagnostic performance regardless of the type of fractures and reader's experience. Further studies are needed to clarify the usefulness of the CNN for the detection of acute rib fractures on CT images in actual clinical practice.

Major trauma with only dynamic criteria: is the routine use of whole-body CT as a first level examination justified?

PURPOSE

Risks and benefits of systematic use of whole-body CT (WBCT) in patients with major trauma when no injury is clinically suspected is still subject of controversy. WBCT allows early identification of potentially evolving lesions, but exposes patients to the risk of high radiation dose and iodine contrast agent. The study aimed to assess if WBCT could be avoided in trauma patients with negative clinical examination.

MATERIALS AND METHODS

This retrospective study included polytrauma patients admitted to the Emergency Department in a six-month period, who had undergone a WBCT scan for major dynamic criteria, with hemodynamic stability, absence of clinical and medical risk factors for major trauma. The patients (n = 233) were divided into two groups according to the absence (n = 152) or presence (n = 81) of clinical suspicion of organ injury. The WBCT results were classified as negative, positive for minor and positive for major lesions.

RESULTS

The average patient age was 44 years. CT scans were completely negative in 111 (47.6%) patients, whose 104 (93.7%) were in the negative clinic group. 122 (52.4%) CT scans were positive, 69 (56.6%) for minor lesions and 53 (43.4%) for major lesions. Among the 48 (39.3%) positive CT scans in patients with negative clinic, only 5 (10.4%) were positive for major lesions. We found a significant difference in the frequency of injuries between the clinically negative and clinically positive patient groups (p < 0.001).

CONCLUSION

A thorough clinical examination associated with a primary radiological evaluation may represent a valid diagnostic approach for trauma with only major dynamic criteria to limit the use of WBCT.

Incidence and underreporting of osseous wrist and hand injuries on whole-body computed tomographies at a level 1 trauma center

BACKGROUND

To investigate the incidence of osseous wrist and hand injuries on whole-body computed tomographies (WBCT) at an urban maximum-care trauma center, to report the number of missed cases in primary radiology reports, and to develop an algorithm for improved detection of these injuries.

METHODS

Retrospective analysis reviewing all WBCT for a period of 8 months for osseous wrist and hand injuries. (1) Reconstruction of hands/wrists in three planes (thickness 1-2 mm) and analysis by a blinded musculoskeletal radiologist. (2) Scanning of primary radiology reports and comparison to the re-evaluation. (3) Calculation of the diagnostic accuracy of WBCT during primary reporting. (4) Search for factors potentially influencing the incidence (trauma mechanism, associated injuries, Glasgow Coma Scale, artifacts). (5) Development of an algorithm to improve the detection rate.

RESULTS

Five hundred six WBCT were included between 01/2020 and 08/2020. 59 (11.7%) WBCT showed 92 osseous wrist or hand injuries. Distal intra-articular radius fractures occurred most frequently (n = 24, 26.1%); 22 patients (37.3%) showed multiple injuries. The sensitivity of WBCT in the detection of wrist and hand fractures during primary evaluation was low with 4 positive cases identified correctly (6.8%; 95% CI 1.9 to 16.5), while the specificity was 100% (95% CI 99.2 to 100.0). Forty-three cases (72.9%) were detected on additional imaging after clinical reassessment. Twelve injuries remained undetected (20.3%). Motorcycle accidents were more common in positive cases (22.0% vs. 10.1%, p = 0.006). 98% of positive cases showed additional fractures of the upper and/or lower extremities, whereas 37% of the patients without osseous wrist and hand injuries suffered such fractures (p < 0.001). The remaining investigated factors did not seem to influence the occurrence.

CONCLUSION

Osseous wrist and hand injuries are present in 11.7% on WBCT after polytrauma. 93.2% of injuries were missed primarily, resulting in a very low sensitivity of WBCT during primary reporting. Motorcycle accidents might predispose for these injuries, and they often cause additional fractures of the extremities. Clinical re-evaluation of patients and secondary re-evaluation of WBCT with preparation of dedicated multiplanar reformations are essential in polytrauma cases to detect osseous injuries of wrist and hand reliably.

TRIAL REGISTRATION

The study was registered prospectively on November 17th, 2020, at the German register for clinical trials (DRKS-ID: DRKS00023589 ).

Radiologist errors by modality, anatomic region, and pathology for 1.6 million exams: what we have learned

PURPOSE

To evaluate the feasibility of adding pathology to recent radiologist error characterization schemes of modality and anatomic region and the potential of this data to more specifically inform peer review and peer learning.

METHODS

Quality assurance data originating from 349 radiologists in a national teleradiology practice were collected for 2019. Interpretive errors were simply categorized as major or minor. Reporting or communication errors were classified as administrative errors. Interpretive errors were then divided by modality, anatomic region and placed into one of 64 pathologic categories.

RESULTS

Out of 1,628,464 studies, the discrepancy rate was 0.5% (8181/1,634,201). The 8181 total errors consisted of 2992 major errors (0.18%) and 5189 minor errors (0.32%). Precisely, 3.1% (257/8181) of total errors were administrative. Of major interpretive errors, 75.5% occurred on CT, with CT abdomen and pelvis accounting for 40.4%. The most common pathologic discrepancy for all exams was in the category of mass, nodule, or adenopathy (1583/8181), the majority of which were minor (1315/1583). The most common pathologic discrepancy for the 2937 major interpretive errors was fracture or dislocation (27%; 793/2937), followed by bleed (10.7%; 315/2937).

CONCLUSION

The addition of error-related pathology to peer review is both feasible and practical and provides a more detailed guide to targeted individual and practice-wide peer learning quality improvement efforts. Future research is needed to determine if there are measurable improvements in detection or interpretation of specific pathologies following error feedback and educational interventions.

Novel Radiographic Indexes for Elbow Stability Assessment: Part A-Cadaveric Validation

INTRODUCTION

Elbow bony stability relies primarily on the high anatomic congruency between the humeral trochlea and the ulnar greater sigmoid notch. No practical tools are available to distinguish different morphotypes of the proximal ulna and herewith predict elbow stability. The aim of this study was to assess inter-observer reproducibility, evaluate diagnostic performance and determine responsiveness to change after simulated coronoid process fracture for three novel elbow radiographic indexes.

METHODS

Ten fresh-frozen cadaver specimens of upper limbs from human donors were available for this study. Three primary indexes were defined, as well as two derived angles: Trochlear Depth Index (TDI); Posterior Coverage Index (PCI); Anterior Coverage Index (ACI); radiographic coverage angle (RCA); olecranon-diaphisary angle (ODA). Each index was first measured on standardized lateral radiographs and subsequently by direct measurement after open dissection. Finally, a type II coronoid fracture (Regan and Morrey classification) was created on each specimen and both radiographic and open measurements were repeated. All measurements were conducted by two orthopaedic surgeons and two dedicated musculoskeletal radiologists.

RESULTS

All three indexes showed good or moderate inter-observer reliability and moderate accuracy and precision when compared to the gold standard (open measurement). A significant change between the radiographic TDI and ACI before and after simulated coronoid fracture was observed [TDI: decrease from 0.45 ± 0.03 to 0.39 ± 0.08 (p = 0.035); ACI: decrease from 1.90 ± 0.17 to 1.58 ± 0.21 (p = 0.001)]. As expected, no significant changes were documented for the PCI. Based on these data, a predictive model was generated, able to identify coronoid fractures with a sensitivity of 80% and a specificity of 100%.

CONCLUSION

New, simple and easily reproducible radiological indexes to describe the congruency of the greater sigmoid notch have been proposed. TDI and ACI change significantly after a simulated coronoid fracture, indicating a good responsiveness of these parameters to a pathological condition. Furthermore, combining TDI and ACI in a regression model equation allowed to identify simulated fractures with high sensitivity and specificity. The newly proposed indexes are, therefore, promising tools to improve diagnostic accuracy of coronoid fractures and show potential to enhance perioperative diagnostic also in cases of elbow instability and stiffness.

LEVEL OF EVIDENCE

Basic science study.

CLINICAL RELEVANCE

The newly proposed indexes are promising tools to improve diagnostic accuracy of coronoid fractures as well as to enhance perioperative diagnostic for elbow instability and stiffness.

Detecting pelvic fracture on 3D-CT using deep convolutional neural networks with multi-orientated slab images

Pelvic fracture is one of the leading causes of death in the elderly, carrying a high risk of death within 1 year of fracture. This study proposes an automated method to detect pelvic fractures on 3-dimensional computed tomography (3D-CT). Deep convolutional neural networks (DCNNs) have been used for lesion detection on 2D and 3D medical images. However, training a DCNN directly using 3D images is complicated, computationally costly, and requires large amounts of training data. We propose a method that evaluates multiple, 2D, real-time object detection systems (YOLOv3 models) in parallel, in which each YOLOv3 model is trained using differently orientated 2D slab images reconstructed from 3D-CT. We assume that an appropriate reconstruction orientation would exist to optimally characterize image features of bone fractures on 3D-CT. Multiple YOLOv3 models in parallel detect 2D fracture candidates in different orientations simultaneously. The 3D fracture region is then obtained by integrating the 2D fracture candidates. The proposed method was validated in 93 subjects with bone fractures. Area under the curve (AUC) was 0.824, with 0.805 recall and 0.907 precision. The AUC with a single orientation was 0.652. This method was then applied to 112 subjects without bone fractures to evaluate over-detection. The proposed method successfully detected no bone fractures in all except 4 non-fracture subjects (96.4%).

A scalable physician-level deep learning algorithm detects universal trauma on pelvic radiographs

Pelvic radiograph (PXR) is essential for detecting proximal femur and pelvis injuries in trauma patients, which is also the key component for trauma survey. None of the currently available algorithms can accurately detect all kinds of trauma-related radiographic findings on PXRs. Here, we show a universal algorithm can detect most types of trauma-related radiographic findings on PXRs. We develop a multiscale deep learning algorithm called PelviXNet trained with 5204 PXRs with weakly supervised point annotation. PelviXNet yields an area under the receiver operating characteristic curve (AUROC) of 0.973 (95% CI, 0.960-0.983) and an area under the precision-recall curve (AUPRC) of 0.963 (95% CI, 0.948-0.974) in the clinical population test set of 1888 PXRs. The accuracy, sensitivity, and specificity at the cutoff value are 0.924 (95% CI, 0.912-0.936), 0.908 (95% CI, 0.885-0.908), and 0.932 (95% CI, 0.919-0.946), respectively. PelviXNet demonstrates comparable performance with radiologists and orthopedics in detecting pelvic and hip fractures.

Automatic detection and classification of rib fractures based on patients' CT images and clinical information via convolutional neural network

OBJECTIVE

To develop a convolutional neural network (CNN) model for the automatic detection and classification of rib fractures in actual clinical practice based on cross-modal data (clinical information and CT images).

MATERIALS

In this retrospective study, CT images and clinical information (age, sex and medical history) from 1020 participants were collected and divided into a single-centre training set (n = 760; age: 55.8 ± 13.4 years; men: 500), a single-centre testing set (n = 134; age: 53.1 ± 14.3 years; men: 90), and two independent multicentre testing sets from two different hospitals (n = 62, age: 57.97 ± 11.88, men: 41; n = 64, age: 57.40 ± 13.36, men: 35). A Faster Region-based CNN (Faster R-CNN) model was applied to integrate CT images and clinical information. Then, a result merging technique was used to convert 2D inferences into 3D lesion results. The diagnostic performance was assessed on the basis of the receiver operating characteristic (ROC) curve, free-response ROC (fROC) curve, precision, recall (sensitivity), F1-score, and diagnosis time. The classification performance was evaluated in terms of the area under the ROC curve (AUC), sensitivity, and specificity.

RESULTS

The CNN model showed improved performance on fresh, healing, and old fractures and yielded good classification performance for all three categories when both clinical information and CT images were used compared to the use of CT images alone. Compared with experienced radiologists, the CNN model achieved higher sensitivity (mean sensitivity: 0.95 > 0.77, 0.89 > 0.61 and 0.80 > 0.55), comparable precision (mean precision: 0.91 > 0.87, 0.84 > 0.77, and 0.95 > 0.70), and a shorter diagnosis time (average reduction of 126.15 s).

CONCLUSIONS

A CNN model combining CT images and clinical information can automatically detect and classify rib fractures with good performance and feasibility in actual clinical practice.

KEY POINTS

• The developed convolutional neural network (CNN) performed better in fresh, healing, and old fractures and yielded a good classification performance in three categories, if both (clinical information and CT images) were used compared to CT images alone. • The CNN model had a higher sensitivity and matched precision in three categories than experienced radiologists with a shorter diagnosis time in actual clinical practice.

Clinician and computer: a study on patient perceptions of artificial intelligence in skeletal radiography

BACKGROUND

Up to half of all musculoskeletal injuries are investigated with plain radiographs. However, high rates of image interpretation error mean that novel solutions such as artificial intelligence (AI) are being explored.

OBJECTIVES

To determine patient confidence in clinician-led radiograph interpretation, the perception of AI-assisted interpretation and management, and to identify factors which might influence these views.

METHODS

A novel questionnaire was distributed to patients attending fracture clinic in a large inner-city teaching hospital. Categorical and Likert scale questions were used to assess participant demographics, daily electronics use, pain score and perceptions towards AI used to assist in interpretation of their radiographs, and guide management.

RESULTS

216 questionnaires were included (M=126, F=90). Significantly higher confidence in clinician rather than AI-assisted interpretation was observed (clinician=9.20, SD=1.27 vs AI=7.06, SD=2.13), 95.4% reported favouring clinician over AI-performed interpretation in the event of disagreement.Small positive correlations were observed between younger age/educational achievement and confidence in AI-assistance. Students demonstrated similarly increased confidence (8.43, SD 1.80), and were over-represented in the minority who indicated a preference for AI-assessment over their clinicians (50%).

CONCLUSIONS

Participant's held the clinician's assessment in the highest regard and expressed a clear preference for it over the hypothetical AI assessment. However, robust confidence scores for the role of AI-assistance in interpreting skeletal imaging suggest patients view the technology favourably.Findings indicate that younger, more educated patients are potentially more comfortable with a role for AI-assistance however further research is needed to overcome the small number of responses on which these observations are based.

Diagnostic error in the emergency department: learning from national patient safety incident report analysis

BACKGROUND

Diagnostic error occurs more frequently in the emergency department than in regular in-patient hospital care. We sought to characterise the nature of reported diagnostic error in hospital emergency departments in England and Wales from 2013 to 2015 and to identify the priority areas for intervention to reduce their occurrence.

METHODS

A cross-sectional mixed-methods design using an exploratory descriptive analysis and thematic analysis of patient safety incident reports. Primary data were extracted from a national database of patient safety incidents. Reports were filtered for emergency department settings, diagnostic error (as classified by the reporter), from 2013 to 2015. These were analysed for the chain of events, contributory factors and harm outcomes.

RESULTS

There were 2288 cases of confirmed diagnostic error: 1973 (86%) delayed and 315 (14%) wrong diagnoses. One in seven incidents were reported to have severe harm or death. Fractures were the most common condition (44%), with cervical-spine and neck of femur the most frequent types. Other common conditions included myocardial infarctions (7%) and intracranial bleeds (6%). Incidents involving both delayed and wrong diagnoses were associated with insufficient assessment, misinterpretation of diagnostic investigations and failure to order investigations. Contributory factors were predominantly human factors, including staff mistakes, healthcare professionals' inadequate skillset or knowledge and not following protocols.

CONCLUSIONS

Systems modifications are needed that provide clinicians with better support in performing patient assessment and investigation interpretation. Interventions to reduce diagnostic error need to be evaluated in the emergency department setting, and could include standardised checklists, structured reporting and technological investigation improvements.

Comparison of X-Ray Imaging and Computed Tomography Scan in the Evaluation of Knee Trauma

Background and objectives: The aim of the study was to compare the accuracy of X-ray (XR) imaging according to computed tomography (CT) scanning in the diagnosis of knee bone fractures, and in the determination of fracture characteristics, and to identify CT scan indications in patients with knee trauma. Materials and methods: The patients who presented to the emergency department (ED) due to knee trauma between January 2017 and December 2018 and who underwent XR imaging and CT scans were included in the study. XR images were reinterpreted by an emergency physician. The official reports, which had been interpreted by a radiologist in the hospital automation system for CT images, were considered valid. Results: Five hundred and forty-eight patients were included in the study. Of the patients, 200 (36.5%) had fractures in XR imaging and 208 (38.0%) had fractures in CT scans. Compared to CT scanning, XR imaging was found to have 89% sensitivity, 95% specificity, 92% positive predictive value, and 92% negative predictive value in identifying the fracture. The sensitivity of XR imaging in identifying growth plate fracture, angulation, stepping off, and extension of the fracture into the joint space was determined as 78% and less. According to the kappa value, there was determined a perfect concordance between the XR imaging and CT scans in angulation, stepping off, and extension of the fracture into the joint space. This concordance was moderate in growth plate fractures. Conclusions: XR imaging has a low sensitivity in identifying knee fractures. There is a moderate concordance between XR imaging and CT scanning in identifying growth plate fractures. Therefore, CT scanning should be performed in patients whose fracture type and fracture characteristics are not able to be determined exactly with XR imaging in knee injury.