Deep learning to detect left ventricular structural abnormalities in chest X-rays.

BACKGROUND AND AIMS: Early identification of cardiac structural abnormalities indicative of heart failure is crucial to improving patient outcomes. Chest X-rays (CXRs) are routinely conducted on a broad population of patients, presenting an opportunity to build scalable screening tools for structural abnormalities indicative of Stage B or worse heart failure with deep learning methods. In this study, a model was developed to identify severe left ventricular hypertrophy (SLVH) and dilated left ventricle (DLV) using CXRs. METHODS: A total of 71 589 unique CXRs from 24 689 different patients completed within 1 year of echocardiograms were identified. Labels for SLVH, DLV, and a composite label indicating the presence of either were extracted from echocardiograms. A deep learning model was developed and evaluated using area under the receiver operating characteristic curve (AUROC). Performance was additionally validated on 8003 CXRs from an external site and compared against visual assessment by 15 board-certified radiologists. RESULTS: The model yielded an AUROC of 0.79 (0.76-0.81) for SLVH, 0.80 (0.77-0.84) for DLV, and 0.80 (0.78-0.83) for the composite label, with similar performance on an external data set. The model outperformed all 15 individual radiologists for predicting the composite label and achieved a sensitivity of 71% vs. 66% against the consensus vote across all radiologists at a fixed specificity of 73%. CONCLUSIONS: Deep learning analysis of CXRs can accurately detect the presence of certain structural abnormalities and may be useful in early identification of patients with LV hypertrophy and dilation. As a resource to promote further innovation, 71 589 CXRs with adjoining echocardiographic labels have been made publicly available.

Lower Extremity Growth according to AI Automated Femorotibial Length Measurement on Slot-Scanning Radiographs in Pediatric Patients.

Background Commonly used pediatric lower extremity growth standards are based on small, dated data sets. Artificial intelligence (AI) enables creation of updated growth standards. Purpose To train an AI model using standing slot-scanning radiographs in a racially diverse data set of pediatric patients to measure lower extremity length and to compare expected growth curves derived using AI measurements to those of the conventional Anderson-Green method. Materials and Methods This retrospective study included pediatric patients aged 0-21 years who underwent at least two slot-scanning radiographs in routine clinical care between August 2015 and February 2022. A Mask Region-based Convolutional Neural Network was trained to segment the femur and tibia on radiographs and measure total leg, femoral, and tibial length; accuracy was assessed with mean absolute error. AI measurements were used to create quantile polynomial regression femoral and tibial growth curves, which were compared with the growth curves of the Anderson-Green method for coverage based on the central 90% of the estimated growth distribution. Results In total, 1874 examinations in 523 patients (mean age, 12.7 years ± 2.8 [SD]; 349 female patients) were included; 40% of patients self-identified as White and not Hispanic or Latino, and the remaining 60% self-identified as belonging to a different racial or ethnic group. The AI measurement training, validation, and internal test sets included 114, 25, and 64 examinations, respectively. The mean absolute errors of AI measurements of the femur, tibia, and lower extremity in the test data set were 0.25, 0.27, and 0.33 cm, respectively. All 1874 examinations were used to generate growth curves. AI growth curves more accurately represented lower extremity growth in an external test set (n = 154 examinations) than the Anderson-Green method (90% coverage probability: 86.7% [95% CI: 82.9, 90.5] for AI model vs 73.4% [95% CI: 68.4, 78.3] for Anderson-Green method; χ2 test, P < .001). Conclusion Lower extremity growth curves derived from AI measurements on standing slot-scanning radiographs from a diverse pediatric data set enabled more accurate prediction of pediatric growth. © RSNA, 2024 Supplemental material is available for this article.

Applications of Diffusion-Weighted MRI to the Musculoskeletal System.

Diffusion-weighted imaging (DWI) is an established MRI technique that can investigate tissue microstructure at the scale of a few micrometers. Musculoskeletal tissues typically have a highly ordered structure to fulfill their functions and therefore represent an optimal application of DWI. Even more since disruption of tissue organization affects its biomechanical properties and may indicate irreversible damage. The application of DWI to the musculoskeletal system faces application-specific challenges on data acquisition including susceptibility effects, the low T2 relaxation time of most musculoskeletal tissues (2-70 msec) and the need for sub-millimetric resolution. Thus, musculoskeletal applications have been an area of development of new DWI methods. In this review, we provide an overview of the technical aspects of DWI acquisition including diffusion-weighting, MRI pulse sequences and different diffusion regimes to study tissue microstructure. For each tissue type (growth plate, articular cartilage, muscle, bone marrow, intervertebral discs, ligaments, tendons, menisci, and synovium), the rationale for the use of DWI and clinical studies in support of its use as a biomarker are presented. The review describes studies showing that DTI of the growth plate has predictive value for child growth and that DTI of articular cartilage has potential to predict the radiographic progression of joint damage in early stages of osteoarthritis. DTI has been used extensively in skeletal muscle where it has shown potential to detect microstructural and functional changes in a wide range of muscle pathologies. DWI of bone marrow showed to be a valuable tool for the diagnosis of benign and malignant acute vertebral fractures and bone metastases. DTI and diffusion kurtosis have been investigated as markers of early intervertebral disc degeneration and lower back pain. Finally, promising new applications of DTI to anterior cruciate ligament grafts and synovium are presented. The review ends with an overview of the use of DWI in clinical routine. EVIDENCE LEVEL: 5 TECHNICAL EFFICACY: Stage 3.

Generalizing Diffusion Tensor Imaging of the Physis and Metaphysis.

BACKGROUND: Current methods to predict height potential are inaccurate. Predicting height by using MRI of the physeal cartilage has shown promise but the applicability of this technique in different imaging setups has not been well-evaluated. PURPOSE: To assess variability in diffusion tensor imaging of the physis and metaphysis (DTI-P/M) of the distal femur between different scanners, imaging parameters, tractography software, and resolution. STUDY TYPE: Prospective. POPULATION/SUBJECTS: Eleven healthy subjects (five males and six females ages 10-16.94). FIELD STRENGTH/SEQUENCE: 3 T; DTI single shot echo planar sequences. ASSESSMENT: Physeal DTI tract measurements of the distal femur were compared between different scanners, imaging parameters, tractography settings, interpolation correction, and tractography software. STATISTICAL TESTS: Bland-Altman, Spearman correlation, linear regression, and Shapiro-Wilk tests. Threshold for statistical significance was set at P = 0.05. RESULTS: DTI tract values consistently showed low variability with different imaging and analysis settings. Vendor to vendor comparison exhibited strong correlation (ρ = 0.93) and small but significant bias (bias -5.76, limits of agreement [LOA] -24.31 to 12.78). Strong correlation and no significant difference were seen between technical replicates of the General Electric MRI scanner (ρ = 1, bias 0.17 [LOA -1.5 to 1.2], P = 0.42) and the Siemens MRI scanner (ρ = 0.89, bias = 0.56, P = 0.71). Different voxel sizes (1 × 1 × 2 mm3 vs. 2 × 2 × 3 mm3) did not significantly affect DTI values (bias = 1.4 [LOA -5.7 to 8.4], P = 0.35) but maintained a strong correlation (ρ = 0.82). Gap size (0 mm vs. 0.6 mm) significantly affects tract volume (bias = 1.8 [LOA -5.4 to 1.8]) but maintains a strong correlation (ρ = 0.93). Comparison of tractography algorithms generated significant differences in tract number, length, and volume while maintaining correlation (ρ = 0.86, 0.99, 0.93, respectively). Comparison of interobserver variability between different tractography software also revealed significant differences while maintaining high correlation (ρ = 0.85-0.98). DATA CONCLUSION: DTI of the pediatric physis cartilage shows high reproducibility between different imaging and analytic parameters. EVIDENCE LEVEL: 2 TECHNICAL EFFICACY: Stage 1.

UTE T2* cartilage mapping in the hip: a pilot study assessing cartilage in patients with femoroacetabular impingement.

BACKGROUND: UTE T2* cartilage mapping use in patients undergoing femoroacetabular impingement (FAI) has been lacking but may allow the detection of early cartilage damage. PURPOSE: To assess the reproducibility of UTE T2* cartilage mapping and determine the difference in UTE T2* values between FAI and asymptomatic patients and to evaluate the correlation between UTE T2* values and patient-reported symptoms. MATERIAL AND METHODS: Prospective evaluation of both hips (7 FAI and 7 asymptomatic patients). Bilateral hip 3-T MRI scans with UTE T2* cartilage maps were acquired. A second MRI scan was acquired 1-9 months later. Cartilage was segmented into anterosuperior, superior, and posterosuperior regions. Assessment was made of UTE T2* reproducibility (ICC). Mean UTE T2* values in patients were compared (t-tests) and correlation was made with patient-reported outcomes (Spearman's). RESULTS: ICCs of mean UTE T2* were as follows: acetabular, 0.82 (95% CI=0.50-0.95); femoral, 0.76 (95% CI=0.35-0.92). Significant strong correlation was found between mean acetabular UTE T2* values and iHOT12 (ρ = -0.63) and moderate correlation with mHHS (ρ = -0.57). There was no difference in mean UTE T2* values between affected vs. non-affected FAI hips. FAI-affected hips had significantly higher values in acetabulum vs. asymptomatic patients (13.47 vs. 12.55 ms). There was no difference in mean femoral cartilage values between the FAI-affected hips vs. asymptomatic patients. The posterosuperior femoral region had a higher mean value in non-affected FAI hips vs. asymptomatic patients (12.60 vs. 11.53 ms). CONCLUSION: UTE T2* cartilage mapping had excellent reproducibility. Affected FAI hips had higher mean acetabular UTE T2* values than asymptomatic patients. Severity of patient-reported symptoms correlates with UTE T2* acetabular cartilage values.

Acute patellar dislocation: how skeletal maturity affects patterns of injury.

OBJECTIVE: The main objective of this study was to understand the role of skeletal maturity in the different patterns of osteochondral and ligamentous injuries after an acute lateral patellar dislocation. MATERIALS AND METHODS: Two radiologists independently reviewed MRIs of 212 knees performed after an acute lateral patellar dislocation to evaluate the presence of high-grade patellar osteochondral injury, femoral osteochondral injury, and medial patellofemoral ligament injury. The association of skeletal maturity (indicated by a closed distal femoral physis), age, sex, and first-time versus recurrent dislocation with each of these various lesions was analyzed using Chi-square or T test, and multivariable logistic regression with estimation of odds ratios (OR). RESULTS: Skeletal maturity was significantly associated with high-grade patellar osteochondral injury [OR=2.72 (95% CI 1.00, 7.36); p=0.049] and femoral-side MPFL tear [OR=2.34 (95% CI 1.05, 5.25); p=0.039]. Skeletal immaturity was significantly associated with patellar-side MPFL tear [OR=0.35 (95% CI 0.14, 0.90); p=0.029]. CONCLUSION: Patterns of injury to the patella and medial patellofemoral ligament vary notably between the skeletally immature and mature, and these variations may be explained by the inherent weakness of the patellar secondary physis.

Stressed or fractured: MRI differentiating indicators of physeal injury.

OBJECTIVE: To identify MRI findings that can indicate chronic physeal stress injury and differentiate it from acute Salter-Harris (SH) fracture of the pediatric knee or wrist. METHODS: IRB-approved retrospective study of consecutively selected knee and wrist MRIs from 32 athletes with chronic physeal stress injury and 30 children with acute SH fracture. MRI characteristics (physeal patency, physeal thickening, physeal signal intensity (SI), continuity of the zone of provisional calcification (ZPC), integrity of the periosteum and/or perichondrium, pattern of periphyseal and soft tissue edema signal, and joint effusion) were compared. RESULTS: Forty-eight chronic physeal stress injuries (mean age 13.1 years [8.2-17.5 years]) and 35 SH fractures (mean age 13.3 years [5.1-16.0 years]) were included. Any physeal thickening was more common with chronic stress injury (98% vs 77%, p = 0.003). Abnormal physeal SI was more common with SH fractures (91% vs 67%, p = 0.008). ZPC discontinuity strongly suggested chronic stress injury (79% vs 49%, p < 0.004). Periosteal and/or perichondrial elevation or rupture and soft tissue edema characterized most of the acute SH fractures (p < 0.001) and were seen only in 1 chronic stress injury (< 2%). While periphyseal edema was not significantly different in the two groups (p = 0.890), a joint effusion was associated with acute SH fracture (p < 0.001). CONCLUSION: Chronic physeal stress injury of the pediatric knee and wrist shows higher incidence of ZPC discontinuity and focal physeal thickening compared to SH fracture, reflecting disruption in normal endochondral ossification. However, these findings can overlap in the 2 groups. Periosteal and/or perichondrial injury, soft tissue edema signal, and joint effusion strongly suggest SH fracture and are rarely present with chronic stress injury.

MR Imaging of the Developing Pediatric Marrow: An Overview of Pearls and Pitfalls.

The dynamic and developing pediatric skeleton is a well-elucidated process that occurs in a stepwise faction. Normal development has been reliably tracked and described with Magnetic Resonance (MR) imaging. The recognition of the normal patterns of skeletal development is essential, as normal development may mimic pathology and vice versa. The authors review normal skeleton maturation and the corollary imaging findings while highlighting common marrow imaging pitfalls and pathology.

Fibrous hamartoma of infancy: imaging findings.

BACKGROUND: Fibrous hamartoma of infancy is a benign tumor that typically arises within the first 2 years of life in the subcutaneous and lower dermal layers. Diagnosis can be challenging as it is a rare tumor, and the imaging appearance is not well known. OBJECTIVE: To describe the imaging features in 4 cases of fibrous hamartoma of infancy focusing on ultrasound (US) and magnetic resonance (MR) findings. MATERIALS AND METHODS: In this retrospective IRB-approved study, informed consent was waived. We searched patient charts for histopathology-confirmed fibrous hamartoma of infancy diagnosis between November 2013 and November 2022. We found four cases, three boys and one girl, and the mean age was 1.4 years (5 months-3 years). The lesions were located in the axilla, posterior elbow, posterior neck, and lower back. All four patients underwent ultrasound evaluation of the lesion, and two patients also underwent MRI evaluation. The imaging findings were reviewed by consensus by two pediatric radiologists. RESULTS: US imaging revealed subcutaneous lesions with variably defined hyperechoic regions and intervening hypoechoic bands resulting in a linear "serpentine" pattern or a "multiple semicircle" pattern. MR imaging evidenced heterogeneous soft tissue masses, localized in the subcutaneous fat, and showed hyperintense fat interspersed with hypointense septations on both T1- and T2-weighted images. CONCLUSION: Fibrous hamartoma of infancy has a suggestive appearance on US with heterogeneous, echogenic subcutaneous lesions with intervening hypoechoic portions, in parallel or circumferential arrangement that can be seen as a serpentine or semicircular pattern. On MRI, interspersed macroscopic fatty components show high signal intensity on T1- and T2-weighted images and reduced signal on fat-suppressed inversion recovery images, with irregular peripheral enhancement.

Artificial intelligence to identify fractures on pediatric and young adult upper extremity radiographs.

BACKGROUND: Pediatric fractures are challenging to identify given the different response of the pediatric skeleton to injury compared to adults, and most artificial intelligence (AI) fracture detection work has focused on adults. OBJECTIVE: Develop and transparently share an AI model capable of detecting a range of pediatric upper extremity fractures. MATERIALS AND METHODS: In total, 58,846 upper extremity radiographs (finger/hand, wrist/forearm, elbow, humerus, shoulder/clavicle) from 14,873 pediatric and young adult patients were divided into train (n = 12,232 patients), tune (n = 1,307), internal test (n = 819), and external test (n = 515) splits. Fracture was determined by manual inspection of all test radiographs and the subset of train/tune radiographs whose reports were classified fracture-positive by a rule-based natural language processing (NLP) algorithm. We trained an object detection model (Faster Region-based Convolutional Neural Network [R-CNN]; "strongly-supervised") and an image classification model (EfficientNetV2-Small; "weakly-supervised") to detect fractures using train/tune data and evaluate on test data. AI fracture detection accuracy was compared with accuracy of on-call residents on cases they preliminarily interpreted overnight. RESULTS: A strongly-supervised fracture detection AI model achieved overall test area under the receiver operating characteristic curve (AUC) of 0.96 (95% CI 0.95-0.97), accuracy 89.7% (95% CI 88.0-91.3%), sensitivity 90.8% (95% CI 88.5-93.1%), and specificity 88.7% (95% CI 86.4-91.0%), and outperformed a weakly-supervised model (AUC 0.93, 95% CI 0.92-0.94, P < 0.0001). AI accuracy on cases preliminary interpreted overnight was higher than resident accuracy (AI 89.4% vs. 85.1%, 95% CI 87.3-91.5% vs. 82.7-87.5%, P = 0.01). CONCLUSION: An object detection AI model identified pediatric upper extremity fractures with high accuracy.

Subcutaneous fat necrosis of the newborn: a pictorial essay of an under-recognized entity.

Subcutaneous fat necrosis of the newborn is a self-limited disorder predominantly affecting full-term and post-term neonates during the first 6 weeks after birth. Subcutaneous fat necrosis can be focal or multifocal and affect one or both sides with a predilection for areas of pressure in certain anatomical areas. Subcutaneous fat necrosis of the newborn is associated with perinatal asphyxia and other neonatal and maternal risk factors. Subcutaneous fat necrosis of the newborn presents as a self-limited area of dermal edema followed by indurated subcutaneous plaques, or nontender and mobile nodules, sometimes with skin discoloration [1-3]. The diagnosis is based on the child's history and physical examination, but when in doubt, imaging is helpful. US is the imaging modality of choice to confirm the diagnosis of subcutaneous fat necrosis of the newborn because it provides the best resolution of superficial lesions, requires no sedation and lacks ionizing radiation. US can also help evaluate and characterize other pathologies affecting the superficial subcutaneous soft tissues at this age. Familiarity with subcutaneous fat necrosis of the newborn is important to make a prompt and precise diagnosis and avoid unnecessary imaging tests or invasive procedures.

Diffusion tensor imaging of the physis: the ABC's.

The physis, or growth plate, is the primary structure responsible for longitudinal growth of the long bones. Diffusion tensor imaging (DTI) is a technique that depicts the anisotropic motion of water molecules, or diffusion. When diffusion is limited by cellular membranes, information on tissue microstructure can be acquired. Tractography, the visual display of the direction and magnitude of water diffusion, provides qualitative visualization of complex cellular architecture as well as quantitative diffusion metrics that appear to indirectly reflect physeal activity. In the growing bones, DTI depicts the columns of cartilage and new bone in the physeal-metaphyseal complex. In this "How I do It", we will highlight the value of DTI as a clinical tool by presenting DTI tractography of the physeal-metaphyseal complex of children and adolescents during normal growth, illustrating variation in qualitative and quantitative tractography metrics with age and skeletal location. In addition, we will present tractography from patients with physeal dysfunction caused by growth hormone deficiency and physeal injury due to trauma, chemotherapy, and radiation therapy. Furthermore, we will delineate our process, or "DTI pipeline," from image acquisition to data interpretation.

Diffusion Tensor Imaging of the Knee to Predict Childhood Growth.

Background Accurate and precise methods to predict growth remain lacking. Diffusion tensor imaging (DTI) depicts the columnar structure of the physis and metaphyseal spongiosa and provides measures of tract volume and length that may help predict growth. Purpose To validate physeal DTI metrics as predictors of height velocity (1-year height gain from time of MRI examination) and total height gain (height gain from time of MRI examination until growth stops) and compare the prediction accuracy with bone age-based models. Materials and Methods Femoral DTI studies (b values = 0 and 600 sec/mm2; directions = 20) of healthy children who underwent MRI of the knee between February 2012 and December 2016 were retrospectively analyzed. Children with height measured at MRI and either 1 year later (height velocity) or after growth cessation (total height gain, mean = 34 months from MRI) were included. Physeal DTI tract volume and length were correlated with height velocity and total height gain. Multilinear regression was used to assess the potential of DTI metrics in the prediction of both parameters. Bland-Altman plots were used to compare root mean square error (RMSE) and bias in height prediction using DTI versus bone age methods. Results Eighty-nine children (mean age, 13 years ± 3 [SD]; 47 boys) had height velocity measured, and 70 (mean age, 14 years ± 1; 36 girls) had total height gain measured. Tract volumes correlated with height velocity (r2 = 0.49) and total height gain (r2 = 0.46) (P < .001 for both) after controlling for age and sex. Tract volume was the strongest predictor for height velocity and total height gain. An optimal multilinear model including tract volume improved prediction of height velocity (R2 = 0.63, RMSE = 1.7 cm) and total height gain (R2 = 0.59, RMSE = 1.8 cm) compared with bone age-based methods (height velocity: R2 = 0.32, RMSE = 2.9 cm; total height gain: R2 = 0.42, RMSE = 5.0 cm). Conclusion Models using tract volume derived from diffusion tensor imaging may perform better than bone age-based models in children for the prediction of height velocity and total height gain. © RSNA, 2022.

Feasibility of ultrashort echo time (UTE) T2* cartilage mapping in the hip: a pilot study.

BACKGROUND: Ultrashort echo time (UTE) T2* is sensitive to molecular changes within the deep calcified layer of cartilage. Feasibility of its use in the hip needs to be established to determine suitability for clinical use. PURPOSE: To establish feasibility of UTE T2* cartilage mapping in the hip and determine if differences in regional values exist. MATERIAL AND METHODS: MRI scans with UTE T2* cartilage maps were prospectively acquired on eight hips. Hip cartilage was segmented into whole and deep layers in anterosuperior, superior, and posterosuperior regions. Quantitative UTE T2* maps were analyzed (independent one-way ANOVA) and reliability was calculated (ICC). RESULTS: UTE T2* mean values (anterosuperior, superior, posterosuperior): full femoral layer (19.55, 18.43, 16.84 ms) (P=0.004), full acetabular layer (19.37, 17.50, 16.73 ms) (P=0.013), deep femoral layer (18.68, 17.90, 15.74 ms) (P=0.010), and deep acetabular layer (17.81, 16.18, 15.31 ms) (P=0.007). Values were higher in anterosuperior compared to posterosuperior regions (mean difference; 95% confidence interval [CI]): full femur layer (2.71 ms; 95% CI 0.91-4.51: P=0.003), deep femur layer (2.94 ms; 95% CI 0.69-5.19; P=0.009), full acetabular layer (2.63 ms 95% CI 0.55-4.72; P=0.012), and deep acetabular layer (2.50 ms; 95% CI 0.69-4.30; P=0.006). Intra-reader (ICC 0.89-0.99) and inter-reader reliability (ICC 0.63-0.96) were good to excellent for the majority of cartilage layers. CONCLUSION: UTE T2* cartilage mapping was feasible in the hip with mean values in the range of 16.84-19.55 ms in the femur and 16.73-19.37 ms in the acetabulum. Significantly higher values were present in the anterosuperior region compared to the posterosuperior region.

Mentoring for diversity and inclusion in pediatric radiology: nurturing the next generation of physicians from underrepresented minorities.

The increasing recognition of the need for a diverse workforce as a tool for excellence in medicine has fortified the efforts toward recruitment, retention and development of faculty from racial and ethnic minorities. Despite these efforts, individuals of Black, Hispanic, American Indian and Alaska Native, Native Hawaiian and other Pacific Islander backgrounds remain vastly underrepresented in the radiology workforce. The main impediments to increasing their representation are an insufficient pipeline and the long time required to train a pediatric radiologist. A greater representation of minorities can only be achieved through the enduring nurturing of future pediatric radiologists along every step in the professional life cycle of a physician, from high school through fellowship completion. Restructuring of faculty recruitment and faculty development policies is also required. A key component of faculty development and overall wellness is mentorship. Junior faculty, particularly those from racial and ethnic minorities, benefit from the experience, advice and support of more experienced radiologists. Successful mentorship is key to ensuring that staff from underrepresented backgrounds thrive within their institutions and in turn become successful mentors to younger individuals, thus completing a virtuous cycle of minority mentorship.

Normal and abnormal tendons in children: evaluation with ultrasound.

Ultrasound (US) has emerged as an essential diagnostic tool for evaluating the entire musculoskeletal system in children. The spatial resolution of modern US technology offers unparalleled depiction of superficial anatomy, and motion and blood flow are demonstrated in real time allowing for the quick diagnosis of a wide variety of pathologies. US evaluation of tendons and their structure and function represents one of the best applications of musculoskeletal US. This article reviews some of the more common indications for US of the tendons in children. While not an exhaustive list, the anatomy and pathology examples described should help any pediatric radiologist confronted with a case of tendon pain or loss of function.

Inferring pediatric knee skeletal maturity from MRI using deep learning.

PURPOSE: Many children who undergo MR of the knee to evaluate traumatic injury may not undergo a separate dedicated evaluation of their skeletal maturity, and we wished to investigate how accurately skeletal maturity could be automatically inferred from knee MRI using deep learning to offer this additional information to clinicians. MATERIALS AND METHODS: Retrospective data from 894 studies from 783 patients were obtained (mean age 13.1 years, 47% female). Coronal and sagittal sequences that were T1/PD-weighted were included and resized to 224 × 224 pixels. Data were divided into train (n = 673), tune (n = 48), and test (n = 173) sets, and children were separated across sets. The chronologic age was predicted using deep learning approaches based on a long short-term memory (LSTM) model, which took as input DenseNet-121-extracted features from all T1/PD coronal and sagittal slices. Each test case was manually assigned a bone age by two radiology residents using a reference atlas provided by Pennock and Bomar. The patient's age served as ground truth. RESULTS: The error of the model's predictions for chronological age was not significantly different from that of radiology residents (model M.S.E. 1.30 vs. resident 0.99, paired t-test = 1.47, p = 0.14). Pearson correlation between model and resident prediction of chronologic age was 0.96 (p < 0.001). CONCLUSION: A deep learning-based approach demonstrated ability to infer skeletal maturity from knee MR sequences that was not significantly different from resident performance and did so in less than 2% of the time required by a human expert. This may offer a method for automatically evaluating lower extremity skeletal maturity automatically as part of every MR examination.

Pre-treatment MRI of leukaemia and lymphoma in children: are there differences in marrow replacement patterns on T1-weighted images?

OBJECTIVES: To investigate the prevalence and distribution of specific marrow patterns on pre-treatment magnetic resonance imaging (MRI) examinations in children with leukaemia and lymphoma and with respect to the anatomic location. MATERIALS AND METHODS: This retrospective IRB-approved and HIPAA-compliant study included children with leukaemia or lymphoma who underwent pre-treatment MRI examinations over 18 years (between 1 January 1995 and 31 August 2013). Two radiologists blinded to the clinical diagnosis reviewed each study to determine the presence or absence of abnormal marrow signal and, when present, sub-categorised the pattern into diffuse, patchy, or focal abnormal marrow. Chi-square and Fisher's exact tests were used to compare marrow patterns between leukaemia and lymphoma. RESULTS: The study included 50 children (32 males and 18 females; mean age 9.5 ± 5.3 years) with 54 MRI examinations (27 leukaemia and 27 lymphoma) that included 26 spine and 28 non-spine studies. Marrow replacement was present on 43 (80%) studies, significantly more common with leukaemia than with lymphoma (p = 0.039). The diffuse replacement pattern was significantly more common with leukaemia when compared to lymphoma (p < 0.001) and the focal pattern was only observed with lymphoma. In the spine, the diffuse pattern was observed with lymphoma (3/14, 21%). All patients with leukaemia and MRI outside of the spine showed marrow involvement. CONCLUSION: Marrow replacement is common on MRI from children with leukaemia and lymphoma. A diffuse pattern was significantly associated with leukaemia on studies outside of the spine and a focal pattern was only observed with lymphoma, independently of the anatomic location. KEY POINTS: • Bone marrow replacement on pre-treatment MRI examinations in children with leukaemia and lymphoma was observed in 93% (25/27) and 67% (18/27), respectively. • Diffuse pattern of marrow replacement was significantly more common in leukaemia even though this pattern was also observed with lymphoma on the spine MRI studies. • Focal pattern of marrow replacement was present only with lymphoma and not with leukaemia regardless of the anatomic location.

A Pilot Study on Feasibility of Ultrashort Echo Time T2* Cartilage Mapping in the Sacroiliac Joints.

PURPOSE: Assess feasibility of ultrashort echo time (UTE) T2* cartilage mapping in sacroiliac (SI) joints. METHODS: Prospective magnetic resonance imagings with UTE T2* cartilage maps obtained on 20 SI joints in 10 subjects. Each joint was segmented into thirds by 2 radiologists. The UTE T2* maps were analyzed; reliability and differences in UTE T2* values between radiologists were assessed. RESULTS: Mean UTE T2* value was 10.44 ± 0.60 ms. No difference between right/left SI joints (median, 10.52 vs 10.45 ms; P = 0.940), men/women (median, 10.34 vs. 10.57 ms; P = 0.174), or different anatomic regions (median range 10.55-10.69 ms; P = 0.805). Intraclass correlation coefficients were 0.94 to 0.99 (intraobserver) and 0.91 to 0.96 (interobserver). Mean bias ± standard deviation on Bland-Altman was -0.137 ± 0.196 ms (limits of agreement -0.521 and 0.247) without proportional bias (ß = 0.148, P = 0.534). CONCLUSIONS: The UTE T2* cartilage mapping in the SI joints is feasible with high reader reliability.

Extracardiac imaging findings in COVID-19-associated multisystem inflammatory syndrome in children.

BACKGROUND: Coronavirus disease 2019 (COVID-19)-associated multisystem inflammatory syndrome in children (MIS-C) is an emerging syndrome that presents with a Kawasaki-like disease and multiorgan damage in children previously exposed to COVID-19. OBJECTIVE: To review the extracardiac radiologic findings of MIS-C in a group of children and young adults with a confirmed diagnosis of MIS-C. MATERIALS AND METHODS: In a retrospective study from April 1, 2020, to July 31, 2020, we reviewed the imaging studies of 47 children and adolescents diagnosed with MIS-C, 25 females (53%) and 22 males (47%), with an average age of 8.4 years (range 1.3-20 years). Forty-five had chest radiographs, 8 had abdominal radiographs, 13 had abdominal US or MRI, 2 had neck US, and 4 had brain MRI. RESULTS: Thirty-seven of 45 (82%) patients with chest radiographs had findings, with pulmonary opacities being the most common finding (n=27, 60%), most often bilateral and diffuse, followed by peribronchial thickening (n=26, 58%). Eight patients had normal chest radiographs. On abdominal imaging, small-volume ascites was the most common finding (n=7, 54%). Other findings included right lower quadrant bowel wall thickening (n=3, 23%), gallbladder wall thickening (n=3, 23%), and cervical (n=2) or abdominal (n=2) lymphadenopathy. Of the four patients with brain MRI, one had bilateral parieto-occipital abnormalities and another papilledema. CONCLUSION: The diagnosis of MIS-C and its distinction from other pathologies should be primarily based on clinical presentation and laboratory evidence of inflammation because imaging findings are nonspecific. However, it should be considered in the setting of bilateral diffuse pulmonary opacities, peribronchial thickening, right lower quadrant bowel inflammation or unexplained ascites in a child presenting with Kawasaki-like symptoms and a history of COVID-19 infection or recent COVID-19 exposure.