Deep learning to detect anterior cruciate ligament tear on knee MRI: multi-continental external validation.

OBJECTIVES: To develop a deep-learning algorithm for anterior cruciate ligament (ACL) tear detection and to compare its accuracy using two external datasets. METHODS: A database of 19,765 knee MRI scans (17,738 patients) issued from different manufacturers and magnetic fields was used to build a deep learning-based ACL tear detector. Fifteen percent showed partial or complete ACL rupture. Coronal and sagittal fat-suppressed proton density or T2-weighted sequences were used. A Natural Language Processing algorithm was used to automatically label reports associated with each MRI exam. We compared the accuracy of our model on two publicly available external datasets: MRNet, Bien et al, USA (PLoS Med 15:e1002699, 2018); and KneeMRI, Stajduhar et al, Croatia (Comput Methods Prog Biomed 140:151-164, 2017). Receptor operating characteristics (ROC) curves, area under the curve (AUC), sensitivity, specificity, and accuracy were used to evaluate our model. RESULTS: Our neural networks achieved an AUC value of 0.939 for detection of ACL tears, with a sensitivity of 87% (0.875) and a specificity of 91% (0.908). After retraining our model on Bien dataset and Stajduhar dataset, our algorithm achieved AUC of 0.962 (95% CI 0.930-0.988) and 0.922 (95% CI 0.875, 0.962) respectively. Sensitivity, specificity, and accuracy were respectively 85% (95% CI 75-94%, 0.852), 89% (95% CI 82-97%, 0.894), 0.875 (95% CI 0.817-0.933) for Bien dataset, and 68% (95% CI 54-81%, 0.681), 93% (95% CI 89-97%, 0.934), and 0.870 (95% CI 0.821-0.913) for Stajduhar dataset. CONCLUSION: Our algorithm showed high performance in the detection of ACL tears with AUC on two external datasets, demonstrating its generalizability on different manufacturers and populations. This study shows the performance of an algorithm for detecting anterior cruciate ligament tears with an external validation on populations from countries and continents different from the study population. KEY POINTS: • An algorithm for detecting anterior cruciate ligament ruptures was built from a large dataset of nearly 20,000 MRI with AUC values of 0.939, sensitivity of 87%, and specificity of 91%. • This algorithm was tested on two external populations from different other countries: a dataset from an American population and a dataset from a Croatian population. Performance remains high on these two external validation populations (AUC of 0.962 and 0.922 respectively).

Injuries in Skating and Sledding Winter Sports: Patterns and Imaging Findings.

While skiing and snowboarding are amongst the most common winter sports, skating and sledding activities are also popular for competition or recreation. Related injuries following an acute trauma mainly involve head, spine, upper and lower limbs. For elite athletes, overuse injuries represent a significant burden. In skating, lesions can be related to boot structure and design. This article reviews epidemiology, patterns, and imaging findings of common injuries in ice skating, short track speed skating, curling, luge, bobsleigh, and skeleton.

Local clinical diagnostic reference levels for chest and abdomen CT examinations in adults as a function of body mass index and clinical indication: a prospective multicenter study.

OBJECTIVES: To compare institutional dose levels based on clinical indication and BMI class to anatomy-based national DRLs (NDRLs) in chest and abdomen CT examinations and to assess local clinical diagnostic reference levels (LCDRLs). METHODS: From February 2017 to June 2018, after protocol optimization according to clinical indication and body mass index (BMI) class (< 25; ≥ 25), 5310 abdomen and 1058 chest CT series were collected from 5 CT scanners in a Swiss multicenter group. Clinical indication-based institutional dose levels were compared to the Swiss anatomy-based NDRLs. Statistical significance was assessed (p < 0.05). LCDRLs were calculated as the third quartile of the median dose values for each CT scanner. RESULTS: For chest examinations, dose metrics based on clinical indication were always below P75 NDRL for CTDIvol (range 3.9-6.4 vs. 7.0 mGy) and DLP (164.0-211.2 vs. 250 mGycm) in all BMI classes except for DLP in BMI ≥ 25 (248.8-255.4 vs. 250.0 mGycm). For abdomen examinations, they were significantly lower or not different than P50 NDRLs for all BMI classes (3.8-9.0 vs. 10.0 mGy and 192.9-446.8 vs. 470mGycm). The estimated LCDRLs show a drop in CTDIvol (21% for chest and 32% for abdomen, on average) with respect to current DRLs. When considering BMI stratification, the largest LCDRL difference within the same clinical indication is for renal tumor (4.6 mGy for BMI < 25 vs. 10.0 mGy for BMI ≥ 25; - 117%). CONCLUSION: The results suggest the necessity of estimating clinical indication-based DRLs, especially for abdomen examinations. Stratifying per BMI class allows further optimization of the CT doses. KEY POINTS: • Our data show that clinical indication-based DRLs might be more appropriate than anatomy-based DRLs and might help in reducing large variations in dose levels for the same type of examinations. • Stratifying the data per patient-size subgroups (non-overweight, overweight) allows a better optimization of CT doses and therefore the possibility to set LCDRLs based on BMI class. • Institutions who are fostering continuous dose optimization and LDRLs should consider defining protocols based on clinical indication and BMI group, to achieve ALARA.

Automating Quality Control for Structured Standardized Radiology Reports Using Text Analysis.

Radiology reports describe the findings of a radiologist in an imaging examination, produced for another clinician in order to answer to a clinical indication. Sometimes, the report does not fully answer the question asked, despite guidelines for the radiologist. In this article, a system that controls the quality of reports automatically is described. It notably maps the free text onto MeSH terms and checks if the anatomy and disease terms match in the indication and conclusion of a report. The agreement between manual checks of experienced radiologists and the system is high with automatic checks requiring only a fraction of time. Being able to quality control all reports has the potential to improve report quality and thus limit misunderstandings, loosing time for requesting more information and possibly avoid medical mistakes.