BackgroundMagnetic resonance imaging (MRI) of the knee is the preferred method for diagnosing knee injuries. However, interpretation of knee MRI is time-intensive and subject to diagnostic error and variability. An automated system for interpreting knee MRI could prioritize high-risk patients and assist clinicians in making diagnoses. Deep learning methods, in being able to automatically learn layers of features, are well suited for modeling the complex relationships between medical images and their interpretations. In this study we developed a deep learning model for detecting general abnormalities and specific diagnoses (anterior cruciate ligament [ACL] tears and meniscal tears) on knee MRI exams. We then measured the effect of providing the model’s predictions to clinical experts during interpretation.Methods and findingsOur dataset consisted of 1,370 knee MRI exams performed at Stanford University Medical Center between January 1, 2001, and December 31, 2012 (mean age 38.0 years; 569 [41.5%] female patients). The majority vote of 3 musculoskeletal radiologists established reference standard labels on an internal validation set of 120 exams. We developed MRNet, a convolutional neural network for classifying MRI series and combined predictions from 3 series per exam using logistic regression. In detecting abnormalities, ACL tears, and meniscal tears, this model achieved area under the receiver operating characteristic curve (AUC) values of 0.937 (95% CI 0.895, 0.980), 0.965 (95% CI 0.938, 0.993), and 0.847 (95% CI 0.780, 0.914), respectively, on the internal validation set. We also obtained a public dataset of 917 exams with sagittal T1-weighted series and labels for ACL injury from Clinical Hospital Centre Rijeka, Croatia. On the external validation set of 183 exams, the MRNet trained on Stanford sagittal T2-weighted series achieved an AUC of 0.824 (95% CI 0.757, 0.892) in the detection of ACL injuries with no additional training, while an MRNet trained on the rest of the external data achieved an AUC of 0.911 (95% CI 0.864, 0.958). We additionally measured the specificity, sensitivity, and accuracy of 9 clinical experts (7 board-certified general radiologists and 2 orthopedic surgeons) on the internal validation set both with and without model assistance. Using a 2-sided Pearson’s chi-squared test with adjustment for multiple comparisons, we found no significant differences between the performance of the model and that of unassisted general radiologists in detecting abnormalities. General radiologists achieved significantly higher sensitivity in detecting ACL tears (p-value = 0.002; q-value = 0.019) and significantly higher specificity in detecting meniscal tears (p-value = 0.003; q-value = 0.019). Using a 1-tailed t test on the change in performance metrics, we found that providing model predictions significantly increased clinical experts’ specificity in identifying ACL tears (p-value < 0.001; q-value = 0.006). The primary limitations of our study include lack of surgical ground truth and the small size of ...

Key Points Question How does augmentation with a deep learning segmentation model influence the performance of clinicians in identifying intracranial aneurysms from computed tomographic angiography examinations? Findings In this diagnostic study of intracranial aneurysms, a test set of 115 examinations was reviewed once with model augmentation and once without in a randomized order by 8 clinicians. The clinicians showed significant increases in sensitivity, accuracy, and interrater agreement when augmented with neural network model–generated segmentations. Meaning This study suggests that the performance of clinicians in the detection of intracranial aneurysms can be improved by augmentation using deep learning segmentation models.

It is estimated that 166 200 out-of-hospital cardiac events occur each year in the United States, with approximately 60% of these events treated by emergency medical services (EMS). 1 Reported rates of survival following outof-hospital cardiac arrest (OHCA) vary widely, from 0.2% (Detroit [2007]) 2 to 23% (London, England [2005]). 3 Nationwide, the median reported survival rate is 6.4%. 4 The vast majority of patients who survive OHCA are resuscitated at the scene of the cardiac arrest and subsequently transported to the hospital for definitive care. 5,6 Nevertheless, the practice of EMS systems in cases of refractory OHCA vary widely from agency to agency. Although most systems generally follow the basic life support (BLS) and advanced life support(ALS)generalresuscitationguidelines outlined by the American Heart Association, 7 there is widespread variability in their application. In one study, adherence to American Heart Association guidelines for the out-of-hospital care of cardiac arrest was only 40%. 8 During the past 30 years, several research teams have sought to define objective clinical criteria to identify patients who likely will not benefit from rapid transport to the hospital for further resuscitative efforts. 9-18 Despite this research, many EMS systems still urgently transport patients with refractory cardiac arrest to the hospital for continued resuscitative efforts. 19-21 Rapid transport with lights and siren may pose hazards for EMS personnel and the public and should occur only when the risks of high-speed trans-See also pp 1423 and 1462.

The development of deep learning algorithms for complex tasks in digital medicine has relied on the availability of large labeled training datasets, usually containing hundreds of thousands of examples. The purpose of this study was to develop a 3D deep learning model, AppendiXNet, to detect appendicitis, one of the most common life-threatening abdominal emergencies, using a small training dataset of less than 500 training CT exams. We explored whether pretraining the model on a large collection of natural videos would improve the performance of the model over training the model from scratch. AppendiXNet was pretrained on a large collection of YouTube videos called Kinetics, consisting of approximately 500,000 video clips and annotated for one of 600 human action classes, and then fine-tuned on a small dataset of 438 CT scans annotated for appendicitis. We found that pretraining the 3D model on natural videos significantly improved the performance of the model from an AUC of 0.724 (95% CI 0.625, 0.823) to 0.810 (95% CI 0.725, 0.895). The application of deep learning to detect abnormalities on CT examinations using video pretraining could generalize effectively to other challenging cross-sectional medical imaging tasks when training data is limited.Appendicitis is one of the most common life-threatening abdominal emergencies, with lifetime risk of ranging from 5-10% and an annual incidence of 90-140 per 100,000 individuals 1-3 . Treatment, via appendectomy, remains the most frequently performed surgical intervention in the world 4 . Computed tomography (CT) imaging of the abdomen is the primary imaging modality to make the diagnosis of appendicitis, exclude other etiologies of acute abdominal pain, and allow for surgical planning and intervention. Because a delay in the time between CT imaging diagnosis and surgical appendectomy can worsen patient outcomes 5-9 , there is increasing pressure on hospital systems to provide 24-7 access to advanced imaging and to ensure that the results of urgent findings, such as appendicitis, are rapidly and accurately communicated to the referring physician 10,11 . However, providing rapid and accurate diagnostic imaging is increasingly difficult to sustain for many medical systems and radiology groups as utilization has rapidly expanded; for example, the CT utilization rate in the emergency room increased from 41 per 1000 in 2000 to 74 per 1000 in 2010, with abdominal CTs accounting for over half of this volume 12,13 . Development of automated systems could potentially help improve diagnostic accuracy and reduce time to diagnosis, thereby improving the quality and efficiency of patient care.Recent advancements in deep learning have enabled algorithms to automate a wide variety of medical tasks [14][15][16][17][18][19][20] . A key aspect for the success of deep learning models on these tasks is the availability of large labeled datasets of medical images 21 , usually containing hundreds of thousands of examples. However, it is challenging to curate large labeled medical imaging dat...

The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.