Background: Detection of Lisfranc malalignment leading to the instability of the joint, particularly in subtle cases, has been a concern for foot and ankle care providers. X-ray radiographs are the mainstay in the diagnosis of these injuries; thus, improving the performance of clinicians in interpreting radiographs can noticeably affect the quality of health care in these patients. Here we assessed the performance of deep learning algorithms on weightbearing radiographs for detection of Lisfranc joint malalignment in patients with Lisfranc instability. Methods: In a retrospective study, 640 patients with Lisfranc malalignment leading to instability were recruited plus 640 individuals with uninjured feet and healthy Lisfranc joint as the control group. All radiographs were screened by orthopaedic surgeons. Two deep learning models were trained, validated, and tested (in a ratio 80:10:10) using a single-view (anteroposterior) and 3-view (anteroposterior, lateral, oblique) radiographs. The performances of the models were reported as sensitivity, specificity, positive and negative predictive values, accuracy, F score, and area under the curve (AUC). Results: No significant differences were observed between the patients and the controls regarding age, gender, race, and body mass index. The best deep learning algorithm outperformed our human interpreters (<1% vs ~10% misdiagnosis), 94.8% sensitivity, 96.9% specificity, 98.6% accuracy, 95.8% F score, and 99.4% AUC. Conclusion: Deep learning methods have shown promising potential in acting as an assistant interpreter of radiographic images in patients with Lisfranc malalignment. Developing these algorithms can hasten and improve the accuracy of diagnosis and reduce further costs and burdens on the patients and health care system. Level of Evidence: Level III, case-control Machine Learning study.

Category: Ankle; Trauma Introduction/Purpose: Early and accurate detection of ankle fractures is crucial for reducing future complications. Radiographs are the most abundant imaging techniques for assessing fractures. We believe deep learning (DL) methods, through adequately trained deep convolutional neural networks (DCNNs), can assess radiographic images fast and accurate without human intervention. In this study, we aimed to assess the performance of two different DCNNs in detecting ankle fractures using radiographs compared to the ground truth. Methods: In this retrospective study, our DCNNs were trained using radiographs obtained from 1050 patients with ankle fracture and the same number of individuals with otherwise healthy ankles. Inception V3 and Renet50 pre-trained models were used in our algorithms. Danis-Weber classification method was used. Out of 1050, 72 individuals were labeled as occult fractures as they were not detected in the primary radiographic assessment. Using single-view radiographs was compared with 3-views (anteroposterior, mortise, lateral) for training the DCNNs. Results: Our DCNNs showed a better performance using 3-views images versus single-view based on greater values for accuracy, F-score, and area under the curve (AUC). The sensitivity and specificity in detection of ankle fractures using 3-views were 97.5% and 93.9% using Resnet50 compared to 98.7% and 98.6 using inception V3, respectively. Resnet50 missed 3 occult fractures while Inception V3 missed only one case. In cases that detected the fracture, the saliency map showed the location of the fracture ( Figure 1 ). Conclusion: The performance of our DCNNs showed a promising potential that can be considered in developing the currently used image interpretation programs or as a separate assistant to the clinicians to detect ankle fractures faster and more precisely.

Category: Ankle; Trauma Introduction/Purpose: Weightbearing CT (WBCT) scan provides an ability to compare the ankle joints bilaterally in a 3D manner under physiologic load. According to our recent investigations, 3D volume measurement of the syndesmosis, if measured up to 5cm proximal to the tibial plafond, can detect the instability with an accuracy of 90%, sensitivity of 95.8%, and specificity of 83.3%. However, these values can differ based on the knowledge and experience of the human interpreter. Deep learning, as a subset of machine learning, has shown promising potentials in processing and analyzing images and detecting abnormalities within the images using deep convolutional neural networks (DCNN). Herein, we aimed to assess the accuracy, sensitivity, and specificity of 3D volume WBCT evaluation using DCNN algorithms in patients with subtle syndesmotic instability. Methods: In this study 140 bilateral ankle WBCT scans of patients with subtle syndesmotic instability who were diagnosed intraoperatively were allocated to the patient group. The control group comprised 140 bilateral ankle WBCT images of healthy individuals. We utilized inception V3 model for our DCNN. Data augmentation and transfer learning were used; however, the images were not preprocessed in terms of change in size and resolution. The data were divided as 80:10:10 for training, validation, and test subsets, respectively. The outcome of the study was expressed as sensitivity, specificity, F-score, and the area under the curve (AUC). Results: The performance of our DCNN algorithm showed a sensitivity of 99.41%, specificity of 99.34%, F-score of 99.37%, and 99.99% AUC ( Figure 1 ). The change in loss value of the train data was plateaued after 40 iterations. Axial images were the most appropriate images that were used by the algorithm to detect the instability. Conclusion: In this study we observed that using DCNN in the process of WBCT image interpretation for diagnosis of syndesmotic instability, particularly in subtle cases, makes this modality almost perfect with a very small chance of missing a case. Training a DCNN using a greater number of inputs is still recommended to improve the validity and reliability of this method. Providing a heat map will also help clinicians discover the process of decision-making by these algorithms as DCNNs are sometimes called 'black box'.

Category: Midfoot/Forefoot; Trauma Introduction/Purpose: Diagnosis of subtle Lisfranc joint instability, as a commonly missed foot injury, has remained a concern since it can result in future disabilities if inadequately treated. Weightbearing radiographs (WBR) and conventional CT scans are the most frequent methods in healthcare centers all around the world that are used to assess tarsometatarsal injuries, specifically the Lisfranc joint. However, their accuracy in detecting subtle cases varies depending on the experience and expertise of the interpreter as well as the quality of the images. We aimed to evaluate the use of deep learning and deep convolutional neural network (DCNN) in the detection of subtle Lisfranc instability using WBR and CT scans. Our hypothesis was that this method can increase the accuracy and hasten the interpretation using these modalities. Methods: We gathered 200 WBR and 200 CT scans of cases with subtle Lisfranc instability who were diagnosed intraoperatively; 200 WBR and 200 CT scans of patients with otherwise healthy feet were added as the control group. To increase the confidence in the results we implemented saliency maps to visualize the location of the injury as a heat map and exhibit the process of decision-making by the algorithm. The data of the study was expressed as sensitivity, specificity, accuracy, and the area under the curve (AUC). We used Inception DCNN model as the pre-trained DCNN model in this study. Results: The performance of the DCNN using WBR resulted in sensitivity=93.6%, specificity=91.1%, Accuracy= 94.7, AUC=98.2%. DCNN applied on CT scan resulted in sensitivity=95.8%, specificity=96.9%, accuracy= 93.2, and AUC=98.4%. In cases that the injury was detected correctly by the DCNN, the saliency map had shown the location of the injury correctly as well (100%, Figure 1 ). Conclusion: Here we showed that using DCNN on the currently used interpretation method can significantly improve the accuracy of interpretation using WBR and CT scans in the detection of subtle Lisfranc instability. WBR has lower costs and a lower rate of radiation, thus, improving its performance using deep learning methods can lead to a significant improvement in healthcare quality for the patient and reduced costs for the system.