Scaphoid fractures are the most common carpal fracture,¹ with the highest prevalence in active, young adult men.² Although the estimated annual incidence is 5 scaphoid fractures per 10 000 people,¹ the actual incidence is likely higher, as physicians fail to detect as many as 20% of scaphoid fractures on the initial radiograph.^3,4 Failure to detect and treat scaphoid fractures can lead to detrimental consequences for patients.² Occult scaphoid fractures, ie, scaphoid fractures that are difficult to detect on radiographs, are increasingly recognized as the etiology responsible for a large proportion of scaphoid nonunions.⁵ Patients with scaphoid nonunions are susceptible to further complications, including degenerative wrist arthritis, chronic wrist pain, and carpal collapse.² As scaphoid fractures have the greatest prevalence in the younger working population, socioeconomic burden and productivity loss associated with scaphoid fractures are substantial; the yearly estimated economic burden for nonoperative scaphoid fracture management in the United States is approximately $58 million.^6,7

In this study, we aimed to develop a deep learning model for identifying bacterial keratitis (BK) and fungal keratitis (FK) by using slit-lamp images. We retrospectively collected slit-lamp images of patients with culture-proven microbial keratitis between 1 January 2010 and 31 December 2019 from two medical centers in Taiwan. We constructed a deep learning algorithm consisting of a segmentation model for cropping cornea images and a classification model that applies different convolutional neural networks (CNNs) to differentiate between FK and BK. The CNNs included DenseNet121, DenseNet161, DenseNet169, DenseNet201, EfficientNetB3, InceptionV3, ResNet101, and ResNet50. The model performance was evaluated and presented as the area under the curve (AUC) of the receiver operating characteristic curves. A gradient-weighted class activation mapping technique was used to plot the heat map of the model. By using 1330 images from 580 patients, the deep learning algorithm achieved the highest average accuracy of 80.0%. Using different CNNs, the diagnostic accuracy for BK ranged from 79.6% to 95.9%, and that for FK ranged from 26.3% to 65.8%. The CNN of DenseNet161 showed the best model performance, with an AUC of 0.85 for both BK and FK. The heat maps revealed that the model was able to identify the corneal infiltrations. The model showed a better diagnostic accuracy than the previously reported diagnostic performance of both general ophthalmologists and corneal specialists.

Background: Retinal vascular diseases, including diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), myopic choroidal neovascularization (mCNV), and branch and central retinal vein occlusion (BRVO/CRVO), are considered vision-threatening eye diseases. However, accurate diagnosis depends on multimodal imaging and the expertise of retinal ophthalmologists. Objective: The aim of this study was to develop a deep learning model to detect treatment-requiring retinal vascular diseases using multimodal imaging. Methods: This retrospective study enrolled participants with multimodal ophthalmic imaging data from 3 hospitals in Taiwan from 2013 to 2019. Eye-related images were used, including those obtained through retinal fundus photography, optical coherence tomography (OCT), and fluorescein angiography with or without indocyanine green angiography (FA/ICGA). A deep learning model was constructed for detecting DME, nAMD, mCNV, BRVO, and CRVO and identifying treatment-requiring diseases. Model performance was evaluated and is presented as the area under the curve (AUC) for each receiver operating characteristic curve. Results: A total of 2992 eyes of 2185 patients were studied, with 239, 1209, 1008, 211, 189, and 136 eyes in the control, DME, nAMD, mCNV, BRVO, and CRVO groups, respectively. Among them, 1898 eyes required treatment. The eyes were divided into training, validation, and testing groups in a 5:1:1 ratio. In total, 5117 retinal fundus photos, 9316 OCT images, and 20,922 FA/ICGA images were used. The AUCs for detecting mCNV, DME, nAMD, BRVO, and CRVO were 0.996, 0.995, 0.990, 0.959, and 0.988, respectively. The AUC for detecting treatment-requiring diseases was 0.969. From the heat maps, we observed that the model could identify retinal vascular diseases. Conclusions: Our study developed a deep learning model to detect retinal diseases using multimodal ophthalmic imaging. Furthermore, the model demonstrated good performance in detecting treatment-requiring retinal diseases.

Background: The aim of this study was to evaluate the efficacy of a deep learning system in pterygium grading and recurrence prediction. Methods: This was a single center, retrospective study. Slit-lamp photographs, from patients with or without pterygium, were collected to develop an algorithm. Demographic data, including age, gender, laterality, grading, and pterygium area, recurrence, and surgical methods were recorded. Complex ocular surface diseases and pseudopterygium were excluded. Performance of the algorithm was evaluated by sensitivity, specificity, F1 score, accuracy, and area under the receiver operating characteristic curve. Confusion matrices and heatmaps were created to help explain the results. Results: A total of 237 eyes were enrolled, of which 176 eyes had pterygium and 61 were non-pterygium eyes. The training set and testing set were comprised of 189 and 48 photographs, respectively. In pterygium grading, sensitivity, specificity, F1 score, and accuracy were 80% to 91.67%, 91.67% to 100%, 81.82% to 94.34%, and 86.67% to 91.67%, respectively. In the prediction model, our results showed sensitivity, specificity, positive predictive value, and negative predictive values were 66.67%, 81.82%, 33.33%, and 94.74%, respectively. Conclusions: Deep learning systems can be useful in pterygium grading based on slit lamp photographs. When clinical parameters involved in the prediction of pterygium recurrence were included, the algorithm showed higher specificity and negative predictive value in prediction.

Accurate segmentation of anatomical structures is vital for medical image analysis. The state-of-the-art accuracy is typically achieved by supervised learning methods, where gathering the requisite expert-labeled image annotations in a scalable manner remains a main obstacle. Therefore, annotation-efficient methods that permit to produce accurate anatomical structure segmentation are highly desirable. In this work, we present Contour Transformer Network (CTN), a one-shot anatomy segmentation method with a naturally built-in human-in-the-loop mechanism. We formulate anatomy segmentation as a contour evolution process and model the evolution behavior by graph convolutional networks (GCNs). Training the CTN model requires only one labeled image exemplar and leverages additional unlabeled data through newly introduced loss functions that measure the global shape and appearance consistency of contours. On segmentation tasks of four different anatomies, we demonstrate that our one-shot learning method significantly outperforms non-learning-based methods and performs competitively to the state-of-the-art fully supervised deep learning methods. With minimal human-in-the-loop editing feedback, the segmentation performance can be further improved to surpass the fully supervised methods.

Background: Patients with diabetic kidney disease (DKD) are at higher risk of hypoglycemia than diabetic patients without DKD. We aimed to investigate the temporal trends of severe hypoglycemia in advanced DKD patients transitioning to dialysis and examine risk factors associated with severe hypoglycemia. We also investigated the association of severe hypoglycemia episodes with one-year mortality after initiation of dialysis in patients with advanced DKD. Methods: Using the Taiwan National Health Insurance Research Database, 46,779 advanced DKD patients transitioning to dialysis (Peritoneal dialysis 4216, hemodialysis 42,563) between 1997 and 2011 were enrolled. We calculated the rates of severe hypoglycemia from 5 years before dialysis until 10 years after dialysis. Cox proportional hazard model was used to examine the risk factors of post end stage renal disease (ESRD) one-year hypoglycemia and post ESRD one-year mortality in advanced DKD patients transitioning to dialysis. Results: We found that 11.5% of advanced DKD patients had at least one episode of severe hypoglycemia the year leading up to dialysis initiation. Multivariate analysis revealed hemodialysis compared with peritoneal dialysis, stroke, use of sulfonylurea, glinide, and insulin were associated with higher risk of severe hypoglycemia one year after transitioning to dialysis. Increased frequency of severe hypoglycemia-related hospitalizations was associated with incrementally higher mortality risk one year after transitioning to dialysis (Pre-ESRD hypoglycemia: Hazard ratios: 1.28 (1.18–1.38, p < 0.001), 1.64 (1.49–1.81, p < 0.001) for one, two hypoglycemia-related hospitalizations, respectively; post-ESRD hypoglycemia: HRs of 1.56 (1.40–1.73, p < 0.001), 1.72 (1.39–2.12, p < 0.001) for one, two hypoglycemia-related hospitalizations, respectively (reference group: no hypoglycemia related hospitalization)). Conclusions: Among advanced DKD patients, we observed a progressive elevated risk of hypoglycemia during the critical dialysis transition period. Increased frequency of severe hypoglycemia-related hospitalizations was associated with higher mortality risk one year after transitioning to dialysis. Further study of glycemic management strategies which prevent hypoglycemia during the critical transition period are warranted.