Fully Automated Diagnosis of Anterior Cruciate Ligament Tears on Knee MR Images by Using Deep Learning.

PURPOSE: To investigate the feasibility of using a deep learning-based approach to detect an anterior cruciate ligament (ACL) tear within the knee joint at MRI by using arthroscopy as the reference standard. MATERIALS AND METHODS: A fully automated deep learning-based diagnosis system was developed by using two deep convolutional neural networks (CNNs) to isolate the ACL on MR images followed by a classification CNN to detect structural abnormalities within the isolated ligament. With institutional review board approval, sagittal proton density-weighted and fat-suppressed T2-weighted fast spin-echo MR images of the knee in 175 subjects with a full-thickness ACL tear (98 male subjects and 77 female subjects; average age, 27.5 years) and 175 subjects with an intact ACL (100 male subjects and 75 female subjects; average age, 39.4 years) were retrospectively analyzed by using the deep learning approach. Sensitivity and specificity of the ACL tear detection system and five clinical radiologists for detecting an ACL tear were determined by using arthroscopic results as the reference standard. Receiver operating characteristic (ROC) analysis and two-sided exact binomial tests were used to further assess diagnostic performance. RESULTS: The sensitivity and specificity of the ACL tear detection system at the optimal threshold were 0.96 and 0.96, respectively. In comparison, the sensitivity of the clinical radiologists ranged between 0.96 and 0.98, while the specificity ranged between 0.90 and 0.98. There was no statistically significant difference in diagnostic performance between the ACL tear detection system and clinical radiologists at P < .05. The area under the ROC curve for the ACL tear detection system was 0.98, indicating high overall diagnostic accuracy. CONCLUSION: There was no significant difference between the diagnostic performance of the ACL tear detection system and clinical radiologists for determining the presence or absence of an ACL tear at MRI.© RSNA, 2019Supplemental material is available for this article.

Comparison of image noise and image quality between full-dose abdominal computed tomography scans reconstructed with weighted filtered back projection and half-dose scans reconstructed with improved sinogram-affirmed iterative reconstruction (SAFIRE*).

PURPOSE: To retrospectively compare the image noise, signal-to-noise ratio (SNR), and subjective image quality between CT images acquired with a dual-source, split-dose imaging protocol reconstructed at full and half doses with weighted filtered back projection (wFBP) and an improved sinogram-affirmed iterative reconstruction algorithm (SAFIRE*). METHODS: Fifty-three consecutive patients underwent contrast-enhanced CT of the abdomen using a standardized dual-source, single energy CT protocol. Half-dose images were retrospectively generated using data from one detector only. Full-dose datasets were reconstructed with wFBP, while half-dose datasets were reconstructed with wFBP and SAFIRE* strengths 1-5. Region of interest analysis was performed to assess SNR and noise. Diagnostic acceptability, subjective noise, and spatial resolution were graded on a 10-point scale by two readers. Statistical analysis was carried out with repeated measures analysis of variance, Wilcoxon signed rank test, and Cohen's κ test. RESULTS: With the increasing strengths of SAFIRE*, a progressive reduction in noise and increase in SNR (p < 0.01) was observed. There was a statistically significant decrease in objective noise and increase in SNR in half-dose SAFIRE* strength 4 and 5 reconstructions compared to full-dose reconstructions using wFBP (p < 0.01). Qualitative analysis revealed a progressive increase in diagnostic acceptability, decrease in subjective noise and increase in spatial resolution for half-dose images reconstructed with the increasing strengths of SAFIRE* (p < 0.01). CONCLUSIONS: Half-dose CT images reconstructed with SAFIRE* at strength 4 and 5 have superior image quality compared to full-dose images reconstructed with wFBP. SAFIRE* potentially allows dose reductions in the order of 50% over wFBP.

Deep Learning Approach for Evaluating Knee MR Images: Achieving High Diagnostic Performance for Cartilage Lesion Detection.

Purpose To determine the feasibility of using a deep learning approach to detect cartilage lesions (including cartilage softening, fibrillation, fissuring, focal defects, diffuse thinning due to cartilage degeneration, and acute cartilage injury) within the knee joint on MR images. Materials and Methods A fully automated deep learning-based cartilage lesion detection system was developed by using segmentation and classification convolutional neural networks (CNNs). Fat-suppressed T2-weighted fast spin-echo MRI data sets of the knee of 175 patients with knee pain were retrospectively analyzed by using the deep learning method. The reference standard for training the CNN classification was the interpretation provided by a fellowship-trained musculoskeletal radiologist of the presence or absence of a cartilage lesion within 17 395 small image patches placed on the articular surfaces of the femur and tibia. Receiver operating curve (ROC) analysis and the κ statistic were used to assess diagnostic performance and intraobserver agreement for detecting cartilage lesions for two individual evaluations performed by the cartilage lesion detection system. Results The sensitivity and specificity of the cartilage lesion detection system at the optimal threshold according to the Youden index were 84.1% and 85.2%, respectively, for evaluation 1 and 80.5% and 87.9%, respectively, for evaluation 2. Areas under the ROC curve were 0.917 and 0.914 for evaluations 1 and 2, respectively, indicating high overall diagnostic accuracy for detecting cartilage lesions. There was good intraobserver agreement between the two individual evaluations, with a κ of 0.76. Conclusion This study demonstrated the feasibility of using a fully automated deep learning-based cartilage lesion detection system to evaluate the articular cartilage of the knee joint with high diagnostic performance and good intraobserver agreement for detecting cartilage degeneration and acute cartilage injury. © RSNA, 2018 Online supplemental material is available for this article .

Comparative Analysis of the Radiology of Osteoporotic Vertebral Fractures in Women and Men: Cross-Sectional and Longitudinal Observations from the Canadian Multicentre Osteoporosis Study (CaMos).

We compared two methods for osteoporotic vertebral fracture (VF) assessment on lateral spine radiographs, the Genant semiquantitative (GSQ) technique and a modified algorithm-based qualitative (mABQ) approach. We evaluated 4465 women and 1771 men aged ≥50 years from the Canadian Multicentre Osteoporosis Study with available X-ray images at baseline. Observer agreement was lowest for grade 1 VFs determined by GSQ. Among physician readers, agreement was greater for VFs diagnosed by mABQ (ranging from 0.62 [95% confidence interval (CI) 0.00-1.00] to 0.88 [0.76-1.00]) than by GSQ (ranging from 0.38 [0.17-0.60] to 0.69 [0.54-0.85]). GSQ VF prevalence (16.4% [95% CI 15.4-17.4]) and incidence (10.2/1000 person-years [9.2; 11.2]) were higher than with the mABQ method (prevalence 6.7% [6.1-7.4] and incidence 6.3/1000 person-years [5.5-7.1]). Women had more prevalent and incident VFs relative to men as defined by mABQ but not as defined by GSQ. Prevalent GSQ VFs were predominantly found in the mid-thoracic spine, whereas prevalent mABQ and incident VFs by both methods co-localized to the junction of the thoracic and lumbar spine. Prevalent mABQ VFs compared with GSQ VFs were more highly associated with reduced adjusted L1 to L4 bone mineral density (BMD) (-0.065 g/cm2 [-0.087 to -0.042]), femoral neck BMD (-0.051 g/cm2 [-0.065 to -0.036]), and total hip BMD (-0.059 g/cm2 [-0.076 to -0.041]). Prevalent mABQ VFs compared with prevalent GSQ were also more highly associated with incident VF by GSQ (odds ratio [OR] = 3.3 [2.2-5.0]), incident VF by mABQ (9.0 [5.3-15.3]), and incident non-vertebral major osteoporotic fractures (1.9 [1.2-3.0]). Grade 1 mABQ VFs, but not grade 1 GSQ VFs, were associated with incident non-vertebral major osteoporotic fractures (OR = 3.0 [1.4-6.5]). We conclude that defining VF by mABQ is preferred to the use of GSQ for clinical assessments. © 2017 American Society for Bone and Mineral Research.

Audit of Atypical Femoral Fractures and a Description of Some of Their Features.

PURPOSE: Atypical femoral fractures (AFF) are recently described events related to osteoporosis and, potentially, a rare result of antiresorptive treatment. METHODS: We set out to audit the diagnosis of AFF in an acute hospital. Charts and radiographs were reviewed retrospectively from patients diagnosed with subtrochanteric femoral fractures according to hospital discharge coding at Vancouver General Hospital (VGH), Canada, from January 2005 to March 2013. RESULTS: A total of 3084 patients were discharged from the hospital with a diagnosis of hip fracture between 2005 and 2013. Of these, 204 were coded as having had subtrochanteric fractures; 178 of the patients thus coded had radiographic evidence of other fracture types-usually intertrochanteric fractures. Eleven patients did not have available radiographs. Of the remaining 193 patients whose radiographs were reviewed, 24 (12.4%) fulfilled the published criteria for AFF. OUR OBSERVATIONS WERE: 1) laterality: 13 of 24 AFF (54.2%) were right-sided; 2) there was only one incomplete AFF in this series: a completed fracture was an inclusion criterion, but 1 patient with an AFF had both that fracture and an incomplete fracture and further foci of periosteal or endosteal foci of new bone (PENB) involving the contralateral femur; 3) radiologists had only diagnosed AFF in only 1 of the 24 patients with characteristic radiographic signs of AFF; 4) all but 1 patient had a focus of periosteal and/or endosteal new bone (PENB) through which the fracture line invariably passed, and in the 1 exception the radiography was too poor to be sure of this but there was a symmetrical contralateral focus of PENB; 5) in 19 of 24 patients there was an adequate image of part of the contralateral femur and of these 12 (63%) had a contralateral focus of PENB situated ±2.5 cm from the index lesion site when measured from the upper aspect of the greater trochanter, and in another patient a prior fracture of the contralateral femur had been treated surgically and it was at a symmetrical contralateral location from the index fracture.; 6) in 3 of the 19 patients multiple foci of PENB were detected on the lateral aspect of the contralateral femur even though the examination was of limited extent; and 7) AFFs were associated with bisphosphonate medication in 75% of the patients studied. CONCLUSIONS: Hospital discharge coding misclassified a great majority of femoral fractures as subtrochanteric. As an essential criteria for diagnosing AFF is their subtrochanteric location, this misclassification impaired our ability to retrospectively search for AFF patients. Radiologists tended not to report AFF when typical radiographic characteristics were present. Bilateral and multifocal disease is of interest in pointing to the diagnosis and in suggesting that the mechanism of injury in respect of these unusual fractures is more complex than simple low-energy trauma.

The Radiology of Osteoporotic Vertebral Fractures Redux.

When a low-energy fracture occurs, then osteoporosis has progressed to the point of bony structural failure. Because vertebral fractures are the commonest type of osteoporotic fracture, the correct identification of them becomes important for diagnosis, risk estimation, and management. However, there are no uniformly agreed criteria for their diagnosis. The purpose of this review was to examine the diagnostic radiological strategies available and suggest a coherent approach to diagnosis. Diagnosis had come to focus on comparative changes in vertebral dimensions. However, it has become apparent that mild reductions in vertebral height are of uncertain implication. The importance of structural damage in diagnosis has become recognized in parallel. Relative reductions in vertebral height may not be a necessary nor sufficient criterion by which to diagnose a fracture.

NASCET percent stenosis semi-automated versus manual measurement on CTA.

PURPOSE: To compare North American Symptomatic Carotid Endarterectomy Trial (NASCET) stenosis values and NASCET grade categorization (mild, moderate, severe) of semi-automated vessel analysis software versus manual measurements on computed tomography angiography (CTA). METHODS: There were four observers. Two independently analyzed 81 carotid artery CTAs using semi-automated vessel analysis software according to a blinded protocol. The software measured the narrowest stenosis in millimeters (mm), distal internal carotid artery (ICA) in mm, and calculated percent stenosis based on NASCET criteria. One of these two observers performed this task twice on each carotid, the second analysis was delayed two months in order to mitigate recall bias. Two other observers manually measured the narrowest stenosis in mm, distal ICA in mm, and calculated NASCET percent stenosis in a blinded fashion. The calculated NASCET stenoses were categorized into mild, moderate, or severe. Chi square and analysis of variance (ANOVA) were used to test for statistical differences. RESULTS: ANOVA did not find a statistically significant difference in the mean percent stenosis when comparing the two manual measurements, the two semi-automated measurements, and the repeat semi-automated. Chi square demonstrated that the distribution of grades of stenosis were statistically different (p<0.05) between the manual and semiautomated grades. Semi-automated vessel analysis tended to underestimate the degree of stenosis compared to manual measurement. CONCLUSION: The mean percentage stenosis determined by semi-automated vessel analysis is not significantly different from manual measurement. However, when the data is categorized into mild, moderate and severe stenosis, there is a significant difference between semi-automated and manual measurements. The semi-automated software tends to underestimate the stenosis grade compared to manual measurement.

Rare frontal lobe intraparenchymal epidermoid cyst with atypical imaging.

Epidermoid cysts are slow-growing benign tumors derived from ectodermal tissue that are hypothesized to have been inwardly displaced from the ectodermal surfaces during embryologic development. These cysts represent 1% to 2% of all intracranial tumors, and occur most commonly in the cerebellopontine angle, parasellar region, and subarachnoid spaces of the basal cisterns. Epidermoid cysts that are exclusively intraparenchymal are very rare, and can be difficult to diagnose as they often do not have classic radiologic findings, and share many similar radiologic features to other tumors such as astrocytomas, arachnoid cysts, dermoid cysts, and cavernomas. The authors present a patient with a rare intraparenchymal epidermoid cyst of the frontal lobe with atypical imaging features.

The Effects of Post-Radiation Education Pamphlet On Self-efficacy in Cancer Patients.

Cancer patients undergoing radiotherapy often experience anxiety and a sense of loss of control immediately after completion of their active treatment. This study aims to evaluate, using a cross-sectional survey, the effects of a post-radiation education pamphlet on the self-efficacy of cancer patients in terms of seeking/understanding medical information, seeking support, and coping with radiation side effects. Forty-eight patients from various age groups with a variety of cancer diagnoses were recruited from the Radiation Medicine Program at Princess Margaret Hospital. Descriptive analysis of 35 survey responses suggests that the pamphlet improved patient self-efficacy in terms of seeking and understanding medical information, seeking support, and coping with radiation side effects. The benefits of the pamphlet were also found to be independent of patients' education level. Despite the small sample size, this study demonstrates the effectiveness of the education pamphlet in improving patient self-efficacy and gives support to the pamphlet's utility in improving the transition care of cancer patients finishing radiotherapy. In parallel, 40 Princess Margaret Hospital Radiation Medicine Program staff members were also asked, via online surveys, about their awareness of the pamphlet, and its usefulness in terms of content, format, and timing of the information provided. More than 90% of the staff indicated that the pamphlet was an effective tool. Only 57% of oncologists were aware of the pamphlet's existence.

Increase in passive stiffness at reduced airway smooth muscle length: potential impact on airway responsiveness.

The amplitude of strain in airway smooth muscle (ASM) produced by oscillatory perturbations such as tidal breathing or deep inspiration (DI) influences the force loss in the muscle and is therefore a key determinant of the bronchoprotective and bronchodilatory effects of these breathing maneuvers. The stiffness of unstimulated ASM (passive stiffness) directly influences the amplitude of strain. The nature of the passive stiffness is, however, not clear. In this study, we measured the passive stiffness of ovine ASM at different muscle lengths (relative to in situ length, which was used as a reference length, L(ref)) and states of adaptation to gain insights into the origin of this muscle property. The results showed that the passive stiffness was relatively independent of muscle length, possessing a constant plateau value over a length range from 0.62 to 1.25 L(ref). Following a halving of ASM length, passive stiffness decreased substantially (by 71%) but redeveloped over time ( approximately 30 min) at the shorter length to reach 65% of the stiffness value at L(ref), provided that the muscle was stimulated to contract at least once over a approximately 30-min period. The redevelopment and maintenance of passive stiffness were dependent on the presence of Ca(2+) but unaffected by latrunculin B, an inhibitor of actin filament polymerization. The maintenance of passive stiffness was also not affected by blocking myosin cross-bridge cycling using a myosin light chain kinase inhibitor or by blocking the Rho-Rho kinase (RhoK) pathway using a RhoK inhibitor. Our results suggest that the passive stiffness of ASM is labile and capable of redevelopment following length reduction. Redevelopment and maintenance of passive stiffness following muscle shortening could contribute to airway hyperresponsiveness by attenuating the airway wall strain induced by tidal breathing and DI.