Automated Rib Fracture Detection on Chest X-Ray Using Contrastive Learning

“Buckle” rib fracture: An artifact following cardio-pulmonary resuscitation detected on postmortem CT

Rib fractures: elusive, but important

Comparison of computed tomography and chest radiography in the detection of rib fractures in abused infants

To review studies on deep learning (DL) models for classification, detection, and segmentation of rib fractures in CT data, to determine their risk of bias (ROB), and to analyse the performance of acute rib fracture detection models. Research articles written in English were retrieved from PubMed, Embase, and Web of Science in April 2023. A study was only included if a DL model was used to classify, detect, or segment rib fractures, and only if the model was trained with CT data from humans. For the ROB assessment, the Quality Assessment of Diagnostic Accuracy Studies tool was used. The performance of acute rib fracture detection models was meta-analysed with forest plots. A total of 27 studies were selected. About 75% of the studies have ROB by not reporting the patient selection criteria, including control patients or using 5-mm slice thickness CT scans. The sensitivity, precision, and F1-score of the subgroup of low ROB studies were 89.60% (95%CI, 86.31%-92.90%), 84.89% (95%CI, 81.59%-88.18%), and 86.66% (95%CI, 84.62%-88.71%), respectively. The ROB subgroup differences test for the F1-score led to a p-value below 0.1. ROB in studies mostly stems from an inappropriate patient and data selection. The studies with low ROB have better F1-score in acute rib fracture detection using DL models. This systematic review will be a reference to the taxonomy of the current status of rib fracture detection with DL models, and upcoming studies will benefit from our data extraction, our ROB assessment, and our meta-analysis.

目的 基于计算机的胸腔X线影像疾病检测和分类目前存在误诊率高，准确率低的问题。本文在视觉Transformer（vision Transformer，ViT）预训练模型的基础上，通过迁移学习方法，实现胸腔X线影像辅助诊断，提高诊断准确率和效率。方法 选用带有卷积神经网络（convolutional neural network，CNN）的ViT模型，其在超大规模自然图像数据集中进行了预训练；通过微调模型结构，使用预训练的ViT模型参数初始化主干网络，并迁移至胸腔X线影像数据集中再次训练，实现疾病多标签分类。结果 在IU X-Ray数据集中对ViT迁移学习前、后模型平均AUC（area under ROC curve）得分进行对比分析实验。结果表明，预训练ViT模型平均AUC得分为0.774，与不使用迁移学习相比提升了0.208。并针对模型结构和数据预处理进行了消融实验，对ViT中的注意力机制进行可视化，进一步验证了模型有效性。最后使用Chest X-Ray14和CheXpert数据集训练微调后的ViT模型，平均AUC得分为0.839和0.806，与对比方法相比分别有0.014～0.031的提升。结论 与其他方法相比，ViT模型胸腔X线影像的多标签分类精确度更高，且迁移学习可以在降低训练成本的同时提升ViT模型的分类性能和泛化性。消融实验与模型可视化表明，包含CNN结构的ViT模型能重点关注有意义的区域，高效获取胸腔X线影像的视觉特征。;Objective The chest X-ray-relevant screening and diagnostic method is essential for radiology nowadays. Most of chest X-ray images interpretation is still restricted by clinical experience and challenged for misdiagnose and missed diagnoses. To detect and identify one or more potential diseases in images automatically，it is beneficial for improving diagnostic efficiency and accuracy using computer-based technique. Compared to natural images，multiple lesions are challenged to be detected and distinguished accurately in a single image because abnormal areas have a small proportion and complex representations in chest X-ray images. Current convolutional neural network（CNN）based deep learning models have been widely used in the context of medical imaging. The structure of the CNN convolution kernel has sensitive to local detail information，and it is possible to extract richer image features. However，the convolution kernel cannot be used to get global information，and the features-extracted are restricted of redundant information like its relevance of background， muscles，and bones. The model’s performance in multi-label classification tasks are affected to a certain extent. At present，the vision Transformer（ViT）model has achieved its priorities in computer vision-related tasks. The ViT can be used to capture information simultaneously and effectively for multiple regions of the entire image. However，it is required to use large-scale dataset training to achieve good performance. Due to some factors like patient privacy and manual annotate costs，the size of the chest X-ray image data set has been limited. To reduce the model's dependence on data scale and improve the performance of multi-label classification，we develop the CNN-based ViT pre-training model in terms of the transfer learning method for diagnosis-assisted of chest X-ray image and multi-label classification. Method The CNN-based ViT model is pre-trained on a huge scale ground truth dataset，and it is used to obtain the initial parameters of the model. The model structure is fine-tuned according to the features of chest X-ray dataset. A 1×1 convolution layer is used to convert the chest X-ray images channels between 1 to 3. The number of output nodes of the linear layer in the classifier is balanced from 1 000 to the number of chest X-ray classification labels，and the Sigmoid is used as an activation function. The parameters of the backbone network are initialized in terms of the pre-trained ViT model parameters，and it is trained in the chest X-ray dataset after that to complete multi-label classification. The experiment is configured of Python3. 7 and PyTorch1. 8 to construct the model and RTX3090 GPU for training. Stochastic gradient descent（SGD）optimizer，binary cross-entropy（BCE）loss function，an initial learning rate of 1E-3，the cosine annealing learning rate decay are used. For training，each image is scaled to a size of 512×512 pixels，and a 224×224 pixels area and it is then cropped in random as the model input，and data augmentation is performed randomly by some of the flipping，perspective transformation， shearing，translation，zooming，and changing brightness. For testing，the chest X-ray image is scaled to 256×256 pixels and center crop a 224×224 area to input the trained model. Result The experiment is performed on the IU X-Ray，which is a small-scale chest X-ray dataset. This model is evaluated in quantitative using the average of area under ROC curve （AUC）scores across all classification labels. The results show that the average AUC score of the pre-trained ViT model is 0. 774. The accuracy and training efficiency of the non-pre-trained ViT model is dropped significantly. The average AUC score is reached to 0. 566 only，which is 0. 208 lower. In addition，the attention mechanism heat map is generated based on the ViT model，which can strengthen the interpretability of the model. A series of ablation experiments are carried out for data augmentation，model structure，and batch size design. The fine-tuned ViT model is trained on the Chest-Ray14 and CheXpert dataset as well. The average AUC score is reached to 0. 839 and 0. 806，which is optimized by 0. 014 and 0. 031. Conclusion A pre-trained ViT model is used for the multi-label classification of chest X-ray images via transfer learning. The experimental results illustrate that the ViT has its stronger multi-label classification performance in chest Xray images，and its attention mechanism is beneficial for lesions precision-focused like the interior of the chest cavity and the heart. Transfer learning is potential to improve the classification performance and model generalization of the ViT in small-scale datasets，and the training cost is reduced greatly. Ablation experiments demonstrate that the incorporated model of CNN and Transformer has its priority beyond single-structure model. Data enhancement and the batch size cutting can improve the performance of the model，but smaller scale of batch is still interlinked to longer training span. To improve the model's ability，we predict that future research direction can be focused on the extraction for complex disease and highlevel semantic information，such as their small lesions，disease location，and severity.

Computed tomography radiation dose optimization for the diagnosis of acute rib fractures in infants investigated for suspected physical abuse: An experimental animal study.

CCE-Net: A rib fracture diagnosis network based on contralateral, contextual, and edge enhanced modules

Rib fractures at postmortem computed tomography (PMCT) validated against the autopsy

The coronavirus disease 2019 (COVID-19) led to a dramatic increase in the number of cases of patients with pneumonia worldwide. In this study, we aimed to develop an AI-assisted multistrategy image enhancement technique for chest X-ray (CXR) images to improve the accuracy of COVID-19 classification. Our new classification strategy consisted of 3 parts. First, the improved U-Net model with a variational encoder segmented the lung region in the CXR images processed by histogram equalization. Second, the residual net (ResNet) model with multidilated-rate convolution layers was used to suppress the bone signals in the 217 lung-only CXR images. A total of 80% of the available data were allocated for training and validation. The other 20% of the remaining data were used for testing. The enhanced CXR images containing only soft tissue information were obtained. Third, the neural network model with a residual cascade was used for the super-resolution reconstruction of low-resolution bone-suppressed CXR images. The training and testing data consisted of 1,200 and 100 CXR images, respectively. To evaluate the new strategy, improved visual geometry group (VGG)-16 and ResNet-18 models were used for the COVID-19 classification task of 2,767 CXR images. The accuracy of the multistrategy enhanced CXR images was verified through comparative experiments with various enhancement images. In terms of quantitative verification, 8-fold cross-validation was performed on the bone suppression model. In terms of evaluating the COVID-19 classification, the CXR images obtained by the improved method were used to train 2 classification models. Compared with other methods, the CXR images obtained based on the proposed model had better performance in the metrics of peak signal-to-noise ratio and root mean square error. The super-resolution CXR images of bone suppression obtained based on the neural network model were also anatomically close to the real CXR images. Compared with the initial CXR images, the classification accuracy rates of the internal and external testing data on the VGG-16 model increased by 5.09% and 12.81%, respectively, while the values increased by 3.51% and 18.20%, respectively, for the ResNet-18 model. The numerical results were better than those of the single-enhancement, double-enhancement, and no-enhancement CXR images. The multistrategy enhanced CXR images can help to classify COVID-19 more accurately than the other existing methods.

Rib fractures are present in 10%-15% of thoracic trauma cases but are often missed on chest radiographs, delaying diagnosis and treatment. Artificial intelligence (AI) may improve detection and triage in emergency settings. This study aims to evaluate diagnostic accuracy, processing speed, and technical feasibility of an artificial intelligence-assisted rib fracture detection system using prospectively collected data within a real-world, high-volume emergency department workflow. We conducted an observational feasibility study with prospective data collection of a faster region-based convolutional neural network-based AI model deployed in the emergency department to analyze 23,251 real-world chest radiographs (22,946 anteroposterior; 305 oblique) from April 1 to July 2, 2023. This study was approved by the Institutional Review Board of MacKay Memorial Hospital (IRB No. 20MMHIS483e). AI operated passively, without influencing clinical decision-making. The reference standard was the final report issued by board-certified radiologists. A subset of discordant cases underwent post hoc computed tomography review for exploratory analysis. AI achieved 74.5% sensitivity (95% CI 0.708-0.780), 93.3% specificity (95% CI 0.930-0.937), 24.2% positive predictive value, and 99.2% negative predictive value. Median inference time was 10.6 seconds versus 3.3 hours for radiologist reports (paired Wilcoxon signed-rank test W=112 987.5, P<.001). The analysis revealed peak imaging demand between 08:00 and 16:00 and Thursday-Saturday evenings. A 14-day graphics processing unit outage underscored the importance of infrastructure resilience. The AI system demonstrated strong technical feasibility for real-time rib fracture detection in a high-volume emergency department setting, with rapid inference and stable performance during prospective deployment. Although the system showed high negative predictive value, the observed false-positive and false-negative rates indicate that it should be considered a supportive screening tool rather than a stand-alone diagnostic solution or a replacement for clinical judgment. These findings support further clinician-in-the-loop studies to evaluate clinical feasibility, workflow integration, and impact on diagnostic decision-making. However, interpretation is limited by reliance on radiology reports as the reference standard and the system's passive, non-interventional deployment.

The research community has recently shown significant interest in designing automated systems to detect coronavirus disease 2019 (COVID-19) using deep learning approaches and chest radiography images. However, state-of-the-art deep learning techniques, especially convolutional neural networks (CNNs), demand more learnable parameters and memory. Therefore, they may not be suitable for real-time diagnosis. Thus, the design of a lightweight CNN model for fast and accurate COVID-19 detection is an urgent need. In this paper, a lightweight CNN model called LW-CORONet is proposed that comprises a sequence of convolution, rectified linear unit (ReLU), and pooling layers followed by two fully connected layers. The proposed model facilitates extracting meaningful features from the chest X-ray (CXR) images with only five learnable layers. The proposed model is evaluated using two larger CXR datasets (Dataset-1: 2250 images and Dataset-2: 15,999 images) and the classification accuracy obtained are 98.67% and 99.00% on Dataset-1 and 95.67% and 96.25% on Dataset-2 for multi-class and binary classification cases, respectively. The results are compared with four contemporary pre-trained CNN models as well as state-of-the-art models. The effect of several hyperparameters: different optimization techniques, batch size, and learning rate have also been investigated. The proposed model demands fewer parameters and requires less memory space. Hence, it is effective for COVID-19 detection and can be utilized as a supplementary tool to assist radiologists in their diagnosis.

To examine the combined effects of image resolution and display luminance on observer performance for detection of abnormalities depicted on posteroanterior chest radiographs. A total of 529 radiographs were displayed on a specially constructed view box at three luminance levels (770, 260, and 85 cd/m(2)) and three resolutions (100-microm, 200-microm, and 400-microm pixels). Each image was reviewed nine times by six radiologists who participated in this study. The abnormalities included nodule, pneumothorax, interstitial disease, alveolar infiltrates, and rib fracture. Negative (normal) radiographs were also included. Receiver operating characteristic curves indicated that the effect of image luminance was greater than that of resolution. The detection of pneumothorax, interstitial disease, and rib fracture showed statistically significant differences (P <. 05) due to luminance. The detection of pneumothorax was the only abnormality with a statistically significant difference due to resolution. There was no evidence that luminance was related to image resolution for any of the abnormalities. At a resolution of 400-microm pixels or higher across the field of view and a luminance of 260 cd/m(2) or more, primary diagnosis with posteroanterior chest radiographs is not likely to be affected by the quality of display.

Rib pathology is uniquely difficult and time-consuming for radiologists to diagnose. AI can reduce radiologist workload and serve as a tool to improve accurate diagnosis. To date, no reviews have been performed synthesizing identification of rib fracture data on AI and its diagnostic performance on X-ray and CT scans of rib fractures and its comparison to physicians. The objectives of this study are to analyze the performance of artificial intelligence in diagnosing rib fracture on X-ray and computed tomography (CT) scan using multiple clinical studies and to compare it to that of physicians findings of rib fracture. A literature search was conducted on PubMed and Embase for articles regarding the use of artificial intelligence for the detection of rib fractures up until July 2024. AI model, number of cases, sensitivity, and comparison to physicians data was collected. A total of 29 studies, comprising 125,364 cases, were included in this review. The pooled sensitivity of AI models was 0.853. Nineteen of these studies compared their results to radiologists, orthopedic surgeons, or anesthesiologists, totalling 61 physicians. Of these 19 studies, the radiologists had a pooled sensitivity of 0.750. The sensitivity of AI in these studies by comparison was 0.840. The results suggest that artificial intelligence has a promising role in detecting rib fractures on X-ray and CT scans. In our interpretation, the performance of artificial intelligence is similar to, or better than, that of physicians, alluding to its encouraging potential in a clinical setting as it may reduce physician workload, improve reading efficiency, and lead to better patient outcomes.

Human studies suggest point-of-care ultrasound (POCUS) is superior to radiographs for diagnosing rib fractures, but its efficacy in veterinary medicine remains unclear. This study aimed to compare the sensitivity and specificity of POCUS and digital radiographs for detecting rib fractures in canine cadavers, using necropsy as the reference standard. Nine canine cadavers were randomly assigned to either a fracture or control group, with fractures created surgically. Blinded evaluations were performed by an expert and novice sonographer, as well as a board-certified radiologist and a novice radiograph interpreter. Sensitivity and specificity for detecting rib fractures were 83% and 99.74% for ultrasound and 82% and 99.22% for radiographs, with no significant difference between modalities. However, the time required to identify rib fractures varied significantly, with ultrasound taking considerably longer than radiograph interpretation. The expert and novice sonographers required an average of 26 and 64 min, respectively, whereas the radiologist and novice radiograph interpreter took 3 and 10 min. These findings suggest that POCUS and radiographs provide comparable accuracy in detecting rib fractures in canine cadavers. Excluding the time required to obtain radiographs, ultrasound takes longer than radiograph interpretation to identify rib fractures. While POCUS remains a valuable diagnostic tool, its practicality in a clinical setting needs further investigation.

Rib fractures account for 39% to 50% of chest trauma cases, with X-ray imaging being the primary diagnostic tool in clinical practice. However, diagnosing rib fractures remains challenging due to multiple factors, including poor patient positioning caused by pain, overlapping rib structures that obscure fractures, and suboptimal X-ray exposure settings that reduce contrast or introduce noise. Studies indicate that X-ray sensitivity for detecting rib fractures ranges from only 12% to 40.7%, with up to 50% of fractures remaining undetected, whereas computed tomography (CT) achieves a detection rate of 39% to 66%. Although CT improves fracture detection, its high cost and radiation exposure limit its routine use in clinical settings. This study collected X-ray images from patients with CT-confirmed rib fractures and trained four convolutional neural networks (CNNs): AlexNet, VGG16, GoogLeNet, and MobileNetV2 to enhance AI-based fracture detection. The optimized CNN models effectively identified suspected rib fractures, achieving a classification accuracy of 0.77 to 0.98. We further applied the YOLOv4 object detection model to precisely locate fractures on X-ray images, with detailed loss function analysis using CIoU for bounding box regression. For comparative purposes, we discuss RetinaNet's Focal Loss (FL = (1- p)^γ CE, with γ=2) as an alternative to YOLOv4's BCE, potentially addressing class imbalance more effectively. Additionally, EfficientDet was explored as a scalable detector to complement YOLOv4, offering multi-scale feature fusion via BiFPN. The final results showed that among 11 X-ray images, the AI model successfully identified the exact fracture locations in 7 cases, demonstrating its reliability and clinical potential. This study integrates deep learning and image recognition techniques, proposing an AI-based solution that balances diagnostic accuracy and clinical feasibility. By leveraging CNNs to enhance X-ray interpretation and YOLOv4 for precise fracture localization, with comparisons to RetinaNet and EfficientDet, this method reduces the burden on physicians, improves diagnostic efficiency, and optimizes medical resource allocation. The findings confirm that this AI model can effectively assist in the initial assessment of rib fractures, providing clinicians with more accurate decisionmaking support and advancing AI-assisted diagnostic tools in healthcare.