MRI Radiogenomics of Pediatric Medulloblastoma: A Multicenter Study

Background Radiogenomics of pediatric medulloblastoma (MB) offers an opportunity for MB risk stratification, which may aid therapeutic decision making, family counseling, and selection of patient groups suitable for targeted genetic analysis. Purpose To develop machine learning strategies that identify the four clinically significant MB molecular subgroups. Materials and Methods In this retrospective study, consecutive pediatric patients with newly diagnosed MB at MRI at 12 international pediatric sites between July 1997 and May 2020 were identified. There were 1800 features extracted from T2- and contrast-enhanced T1-weighted preoperative MRI scans. A two-stage sequential classifier was designed-one that first identifies non-wingless (WNT) and non-sonic hedgehog (SHH) MB and then differentiates therapeutically relevant WNT from SHH. Further, a classifier that distinguishes high-risk group 3 from group 4 MB was developed. An independent, binary subgroup analysis was conducted to uncover radiomics features unique to infantile versus childhood SHH subgroups. The best-performing models from six candidate classifiers were selected, and performance was measured on holdout test sets. CIs were obtained by bootstrapping the test sets for 2000 random samples. Model accuracy score was compared with the no-information rate using the Wald test. Results The study cohort comprised 263 patients (mean age ± SD at diagnosis, 87 months ± 60; 166 boys). A two-stage classifier outperformed a single-stage multiclass classifier. The combined, sequential classifier achieved a microaveraged F1 score of 88% and a binary F1 score of 95% specifically for WNT. A group 3 versus group 4 classifier achieved an area under the receiver operating characteristic curve of 98%. Of the Image Biomarker Standardization Initiative features, texture and first-order intensity features were most contributory across the molecular subgroups. Conclusion An MRI-based machine learning decision path allowed identification of the four clinically relevant molecular pediatric medulloblastoma subgroups. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Chaudhary and Bapuraj in this issue.

MRI Correlates of Ototoxicity in the Auditory Pathway in Children Treated for Medulloblastoma

OBJECTIVE

To assess diffusion and perfusion changes of the auditory pathway in pediatric medulloblastoma patients exposed to ototoxic therapies.

STUDY DESIGN

Retrospective cohort study.

SETTING

A single academic tertiary children's hospital.

PATIENTS

Twenty pediatric medulloblastoma patients (13 men; mean age 12.0 ± 4.8 yr) treated with platinum-based chemotherapy with or without radiation and 18 age-and-sex matched controls were included. Ototoxicity scores were determined using Chang Ototoxicity Grading Scale.

INTERVENTIONS

Three Tesla magnetic resonance was used for diffusion tensor and arterial spin labeling perfusion imaging.

MAIN OUTCOME MEASURES

Quantitative diffusion tensor metrics were extracted from the Heschl's gyrus, auditory radiation, and inferior colliculus. Arterial spin labeling perfusion of the Heschl's gyrus was also examined.

RESULTS

Nine patients had clinically significant hearing loss, or Chang grades more than or equal to 2a; 11 patients had mild/no hearing loss, or Chang grades less than 2a. The clinically significant hearing loss group showed reduced mean diffusivity in the Heschl's gyrus (p = 0.018) and auditory radiation (p = 0.037), and decreased perfusion in the Heschl's gyrus (p = 0.001). Mild/no hearing loss group showed reduced mean diffusivity (p = 0.036) in Heschl's gyrus only, with a decrease in perfusion (p = 0.008). There were no differences between groups in the inferior colliculus. There was no difference in fractional anisotropy between patients exposed to ototoxic therapies and controls.

CONCLUSIONS

Patients exposed to ototoxic therapies demonstrated microstructural and physiological alteration of the auditory pathway. The present study shows proof-of-concept use of diffusion tensor imaging to gauge ototoxicity along the auditory pathway. Future larger cohort studies are needed to assess significance of changes in diffusion tensor imaging longitudinally, and the relationship between these changes and hearing loss severity and longitudinal changes of the developing auditory white matter.

IMG-10. MRI-BASED RADIOMIC PROGNOSTIC MARKERS OF DIFFUSE MIDLINE GLIOMA

BACKGROUND

Diffuse midline gliomas (DMG) are lethal pediatric brain tumors with dismal prognoses. Presently, MRI is the mainstay of disease diagnosis and surveillance. We aimed to identify prognostic image-based radiomics markers of DMG and compare its performance to clinical variables at presentation.

METHODS

104 treatment-naïve DMG MRIs from five centers were used (median age=6.5yrs; 18 males, median OS=11mos). We isolated tumor volumes of T1-post-contrast (T1gad) and T2-weighted (T2) MRI for PyRadiomics high-dimensional feature extraction. 900 features were extracted on each image, including first order statistics, 2D/3D Shape, Gray Level Co-occurrence Matrix, Gray Level Run Length Matrix, Gray Level Size Zone Matrix, Neighboring Gray tone Difference Matrix, and Gray Level Dependence Matrix, as defined by Imaging Biomarker Standardization Initiative. Overall survival (OS) served as outcome. 10-fold cross-validation of LASSO Cox regression was used to predict OS. We analyzed model performance using clinical variable (age at diagnosis and sex) only, radiomics only, and radiomics plus clinical variable. Concordance metric was used to assess the Cox model.

RESULTS

Nine radiomic features were selected from T1gad (2 texture wavelet) and T2 (5 first-order features (1 original, 4 wavelet), 2 texture features (1 wavelet, 1 log-sigma). This model demonstrated significantly higher performance than a clinical model alone (C: 0.68 vs 0.59, p<0.001). Adding clinical features to radiomic features slightly improved prediction, but was not significant (C=0.70, p=0.06).

CONCLUSION

Our pilot study shows a potential role for MRI-based radiomics and machine learning for DMG risk stratification and as image-based biomarkers for clinical therapy trials.

Diffusion tensor magnetic resonance imaging of the optic nerves in pediatric hydrocephalus

OBJECTIVE

While conventional imaging can readily identify ventricular enlargement in hydrocephalus, structural changes that underlie microscopic tissue injury might be more difficult to capture. MRI-based diffusion tensor imaging (DTI) uses properties of water motion to uncover changes in the tissue microenvironment. The authors hypothesized that DTI can identify alterations in optic nerve microstructure in children with hydrocephalus.

METHODS

The authors retrospectively reviewed 21 children (< 18 years old) who underwent DTI before and after neurosurgical intervention for acute obstructive hydrocephalus from posterior fossa tumors. Their optic nerve quantitative DTI metrics of mean diffusivity (MD) and fractional anisotropy (FA) were compared to those of 21 age-matched healthy controls.

RESULTS

Patients with hydrocephalus had increased MD and decreased FA in bilateral optic nerves, compared to controls (p < 0.001). Normalization of bilateral optic nerve MD and FA on short-term follow-up (median 1 day) after neurosurgical intervention was observed, as was near-complete recovery of MD on long-term follow-up (median 1.8 years).

CONCLUSIONS

DTI was used to demonstrate reversible alterations of optic nerve microstructure in children presenting acutely with obstructive hydrocephalus. Alterations in optic nerve MD and FA returned to near-normal levels on short- and long-term follow-up, suggesting that surgical intervention can restore optic nerve tissue microstructure. This technique is a safe, noninvasive imaging tool that quantifies alterations of neural tissue, with a potential role for evaluation of pediatric hydrocephalus.

Deep Learning for Pediatric Posterior Fossa Tumor Detection and Classification: A Multi-Institutional Study

BACKGROUND AND PURPOSE

Posterior fossa tumors are the most common pediatric brain tumors. MR imaging is key to tumor detection, diagnosis, and therapy guidance. We sought to develop an MR imaging-based deep learning model for posterior fossa tumor detection and tumor pathology classification.

MATERIALS AND METHODS

The study cohort comprised 617 children (median age, 92 months; 56% males) from 5 pediatric institutions with posterior fossa tumors: diffuse midline glioma of the pons (n = 122), medulloblastoma (n = 272), pilocytic astrocytoma (n = 135), and ependymoma (n = 88). There were 199 controls. Tumor histology served as ground truth except for diffuse midline glioma of the pons, which was primarily diagnosed by MR imaging. A modified ResNeXt-50-32x4d architecture served as the backbone for a multitask classifier model, using T2-weighted MRIs as input to detect the presence of tumor and predict tumor class. Deep learning model performance was compared against that of 4 radiologists.

RESULTS

Model tumor detection accuracy exceeded an AUROC of 0.99 and was similar to that of 4 radiologists. Model tumor classification accuracy was 92% with an F₁ score of 0.80. The model was most accurate at predicting diffuse midline glioma of the pons, followed by pilocytic astrocytoma and medulloblastoma. Ependymoma prediction was the least accurate. Tumor type classification accuracy and F₁ score were higher than those of 2 of the 4 radiologists.

CONCLUSIONS

We present a multi-institutional deep learning model for pediatric posterior fossa tumor detection and classification with the potential to augment and improve the accuracy of radiologic diagnosis.

PENet-a scalable deep-learning model for automated diagnosis of pulmonary embolism using volumetric CT imaging

Pulmonary embolism (PE) is a life-threatening clinical problem and computed tomography pulmonary angiography (CTPA) is the gold standard for diagnosis. Prompt diagnosis and immediate treatment are critical to avoid high morbidity and mortality rates, yet PE remains among the diagnoses most frequently missed or delayed. In this study, we developed a deep learning model-PENet, to automatically detect PE on volumetric CTPA scans as an end-to-end solution for this purpose. The PENet is a 77-layer 3D convolutional neural network (CNN) pretrained on the Kinetics-600 dataset and fine-tuned on a retrospective CTPA dataset collected from a single academic institution. The PENet model performance was evaluated in detecting PE on data from two different institutions: one as a hold-out dataset from the same institution as the training data and a second collected from an external institution to evaluate model generalizability to an unrelated population dataset. PENet achieved an AUROC of 0.84 [0.82-0.87] on detecting PE on the hold out internal test set and 0.85 [0.81-0.88] on external dataset. PENet also outperformed current state-of-the-art 3D CNN models. The results represent successful application of an end-to-end 3D CNN model for the complex task of PE diagnosis without requiring computationally intensive and time consuming preprocessing and demonstrates sustained performance on data from an external institution. Our model could be applied as a triage tool to automatically identify clinically important PEs allowing for prioritization for diagnostic radiology interpretation and improved care pathways via more efficient diagnosis.

Altered cerebral perfusion in children with Langerhans cell histiocytosis after chemotherapy

BACKGROUND AND PURPOSE

Children with Langerhans cell histiocytosis (LCH) may develop a wide array of neurological symptoms, but associated cerebral physiologic changes are poorly understood. We examined cerebral hemodynamic properties of pediatric LCH using arterial spin-labeling (ASL) perfusion magnetic resonance imaging (MRI).

MATERIALS AND METHODS

A retrospective study was performed in 23 children with biopsy-proven LCH. Analysis was performed on routine brain MRI obtained before or after therapy. Region of interest (ROI) methodology was used to determine ASL cerebral blood flow (CBF) (mL/100 g/min) in the following bilateral regions: angular gyrus, anterior prefrontal cortex, orbitofrontal cortex, dorsal anterior cingulate cortex, and hippocampus. Quantile (median) regression was performed for each ROI location. CBF patterns were compared between pre- and posttreatment LCH patients as well as with age-matched healthy controls.

RESULTS

Significantly reduced CBF was seen in posttreatment children with LCH compared to age-matched controls in angular gyrus (P = .046), anterior prefrontal cortex (P = .039), and dorsal anterior cingulate cortex (P = .023). Further analysis revealed dominant perfusion abnormalities in the right hemisphere. No significant perfusion differences were observed in the hippocampus or orbitofrontal cortex.

CONCLUSION

Perfusion in specific cerebral regions may be consistently reduced in children with LCH, and may represent effects of underlying disease physiology and/or sequelae of chemotherapy. Studies that combine a formal cognitive assessment and hemodynamic data may further provide insight into perfusion deficits associated with the disease and the potential neurotoxic effects in children treated by chemotherapy.

Age-Dependent White Matter Characteristics of the Cerebellar Peduncles from Infancy Through Adolescence

Cerebellum-cerebrum connections are essential for many motor and cognitive functions and cerebellar disorders are prevalent in childhood. The middle (MCP), inferior (ICP), and superior cerebellar peduncles (SCP) are the major white matter pathways that permit communication between the cerebellum and the cerebrum. Knowledge about the microstructural properties of these cerebellar peduncles across childhood is limited. Here, we report on a diffusion magnetic resonance imaging tractography study to describe age-dependent characteristics of the cerebellar peduncles in a cross-sectional sample of infants, children, and adolescents from newborn to 17 years of age (N = 113). Scans were collected as part of clinical care; participants were restricted to those whose scans showed no abnormal findings and whose history and exam had no risk factors for cerebellar abnormalities. A novel automated tractography protocol was applied. Results showed that mean tract-FA increased, while mean tract-MD decreased from infancy to adolescence in all peduncles. Rapid changes were observed in both diffusion measures in the first 24 months of life, followed by gradual change at older ages. The shape of the tract profiles was similar across ages for all peduncles. These data are the first to characterize the variability of diffusion properties both across and within cerebellar white matter pathways that occur from birth through later adolescence. The data represent a rich normative data set against which white matter alterations seen in children with posterior fossa conditions can be compared. Ultimately, the data will facilitate the identification of sensitive biomarkers of cerebellar abnormalities.

Deep Learning for Automated Classification of Inferior Vena Cava Filter Types on Radiographs

PURPOSE

To demonstrate the feasibility and evaluate the performance of a deep-learning convolutional neural network (CNN) classification model for automated identification of different types of inferior vena cava (IVC) filters on radiographs.

MATERIALS AND METHODS

In total, 1,375 cropped radiographic images of 14 types of IVC filters were collected from patients enrolled in a single-center IVC filter registry, with 139 images withheld as a test set and the remainder used to train and validate the classification model. Image brightness, contrast, intensity, and rotation were varied to augment the training set. A 50-layer ResNet architecture with fixed pre-trained weights was trained using a soft margin loss over 50 epochs. The final model was evaluated on the test set.

RESULTS

The CNN classification model achieved a F1 score of 0.97 (0.92-0.99) for the test set overall and of 1.00 for 10 of 14 individual filter types. Of the 139 test set images, 4 (2.9%) were misidentified, all mistaken for other filter types that appear highly similar. Heat maps elucidated salient features for each filter type that the model used for class prediction.

CONCLUSIONS

A CNN classification model was successfully developed to identify 14 types of IVC filters on radiographs and demonstrated high performance. Further refinement and testing of the model is necessary before potential real-world application.

Arterial spin labeling perfusion changes of the frontal lobes in children with posterior fossa syndrome

OBJECTIVE

Posterior fossa syndrome (PFS) is a common complication following the resection of posterior fossa tumors in children. The pathophysiology of PFS remains incompletely elucidated; however, the wide-ranging symptoms of PFS suggest the possibility of widespread cortical dysfunction. In this study, the authors utilized arterial spin labeling (ASL), an MR perfusion modality that provides quantitative measurements of cerebral blood flow without the use of intravenous contrast, to assess cortical blood flow in patients with PFS.

METHODS

A database of medulloblastoma treated at the authors' institution from 2004 to 2016 was retrospectively reviewed, and 14 patients with PFS were identified. Immediate postoperative ASL for patients with PFS and medulloblastoma patients who did not develop PFS were compared. Additionally, in patients with PFS, ASL following the return of speech was compared with immediate postoperative ASL.

RESULTS

On immediate postoperative ASL, patients who subsequently developed PFS had statistically significant decreases in right frontal lobe perfusion and a trend toward decreased perfusion in the left frontal lobe compared with controls. Patients with PFS had statistically significant increases in bilateral frontal lobe perfusion after the resolution of symptoms compared with their immediate postoperative imaging findings.

CONCLUSIONS

ASL perfusion imaging identifies decreased frontal lobe blood flow as a strong physiological correlate of PFS that is consistent with the symptomatology of PFS. This is the first study to demonstrate that decreases in frontal lobe perfusion are present in the immediate postoperative period and resolve with the resolution of symptoms, suggesting a physiological explanation for the transient symptoms of PFS.

Posterior fossa syndrome and increased mean diffusivity in the olivary bodies

OBJECTIVE

Posterior fossa syndrome (PFS) is a common postoperative complication following resection of posterior fossa tumors in children. It typically presents 1 to 2 days after surgery with mutism, ataxia, emotional lability, and other behavioral symptoms. Recent structural MRI studies have found an association between PFS and hypertrophic olivary degeneration, which is detectable as T2 hyperintensity in the inferior olivary nuclei (IONs) months after surgery. In this study, the authors investigated whether immediate postoperative diffusion tensor imaging (DTI) of the ION can serve as an early imaging marker of PFS.

METHODS

The authors retrospectively reviewed pediatric brain tumor patients treated at their institution, Lucile Packard Children's Hospital at Stanford, from 2004 to 2016. They compared the immediate postoperative DTI studies obtained in 6 medulloblastoma patients who developed PFS to those of 6 age-matched controls.

RESULTS

Patients with PFS had statistically significant increased mean diffusivity (MD) in the left ION (1085.17 ± 215.51 vs 860.17 ± 102.64, p = 0.044) and variably increased MD in the right ION (923.17 ± 119.2 vs 873.67 ± 60.16, p = 0.385) compared with age-matched controls. Patients with PFS had downward trending fractional anisotropy (FA) in both the left (0.28 ± 0.06 vs 0.23 ± 0.03, p = 0.085) and right (0.29 ± 0.06 vs 0.25 ± 0.02, p = 0.164) IONs compared with age-matched controls, although neither of these values reached statistical significance.

CONCLUSIONS

Increased MD in the ION is associated with development of PFS. ION MD changes may represent an early imaging marker of PFS.

Deep Learning-Assisted Diagnosis of Cerebral Aneurysms Using the HeadXNet Model

IMPORTANCE

Deep learning has the potential to augment clinician performance in medical imaging interpretation and reduce time to diagnosis through automated segmentation. Few studies to date have explored this topic.

OBJECTIVE

To develop and apply a neural network segmentation model (the HeadXNet model) capable of generating precise voxel-by-voxel predictions of intracranial aneurysms on head computed tomographic angiography (CTA) imaging to augment clinicians' intracranial aneurysm diagnostic performance.

DESIGN, SETTING, AND PARTICIPANTS

In this diagnostic study, a 3-dimensional convolutional neural network architecture was developed using a training set of 611 head CTA examinations to generate aneurysm segmentations. Segmentation outputs from this support model on a test set of 115 examinations were provided to clinicians. Between August 13, 2018, and October 4, 2018, 8 clinicians diagnosed the presence of aneurysm on the test set, both with and without model augmentation, in a crossover design using randomized order and a 14-day washout period. Head and neck examinations performed between January 3, 2003, and May 31, 2017, at a single academic medical center were used to train, validate, and test the model. Examinations positive for aneurysm had at least 1 clinically significant, nonruptured intracranial aneurysm. Examinations with hemorrhage, ruptured aneurysm, posttraumatic or infectious pseudoaneurysm, arteriovenous malformation, surgical clips, coils, catheters, or other surgical hardware were excluded. All other CTA examinations were considered controls.

MAIN OUTCOMES AND MEASURES

Sensitivity, specificity, accuracy, time, and interrater agreement were measured. Metrics for clinician performance with and without model augmentation were compared.

RESULTS

The data set contained 818 examinations from 662 unique patients with 328 CTA examinations (40.1%) containing at least 1 intracranial aneurysm and 490 examinations (59.9%) without intracranial aneurysms. The 8 clinicians reading the test set ranged in experience from 2 to 12 years. Augmenting clinicians with artificial intelligence-produced segmentation predictions resulted in clinicians achieving statistically significant improvements in sensitivity, accuracy, and interrater agreement when compared with no augmentation. The clinicians' mean sensitivity increased by 0.059 (95% CI, 0.028-0.091; adjusted P = .01), mean accuracy increased by 0.038 (95% CI, 0.014-0.062; adjusted P = .02), and mean interrater agreement (Fleiss κ) increased by 0.060, from 0.799 to 0.859 (adjusted P = .05). There was no statistically significant change in mean specificity (0.016; 95% CI, -0.010 to 0.041; adjusted P = .16) and time to diagnosis (5.71 seconds; 95% CI, 7.22-18.63 seconds; adjusted P = .19).

CONCLUSIONS AND RELEVANCE

The deep learning model developed successfully detected clinically significant intracranial aneurysms on CTA. This suggests that integration of an artificial intelligence-assisted diagnostic model may augment clinician performance with dependable and accurate predictions and thereby optimize patient care.

MR Imaging-Based Radiomic Signatures of Distinct Molecular Subgroups of Medulloblastoma

BACKGROUND AND PURPOSE

Distinct molecular subgroups of pediatric medulloblastoma confer important differences in prognosis and therapy. Currently, tissue sampling is the only method to obtain information for classification. Our goal was to develop and validate radiomic and machine learning approaches for predicting molecular subgroups of pediatric medulloblastoma.

MATERIALS AND METHODS

In this multi-institutional retrospective study, we evaluated MR imaging datasets of 109 pediatric patients with medulloblastoma from 3 children's hospitals from January 2001 to January 2014. A computational framework was developed to extract MR imaging-based radiomic features from tumor segmentations, and we tested 2 predictive models: a double 10-fold cross-validation using a combined dataset consisting of all 3 patient cohorts and a 3-dataset cross-validation, in which training was performed on 2 cohorts and testing was performed on the third independent cohort. We used the Wilcoxon rank sum test for feature selection with assessment of area under the receiver operating characteristic curve to evaluate model performance.

RESULTS

Of 590 MR imaging-derived radiomic features, including intensity-based histograms, tumor edge-sharpness, Gabor features, and local area integral invariant features, extracted from imaging-derived tumor segmentations, tumor edge-sharpness was most useful for predicting sonic hedgehog and group 4 tumors. Receiver operating characteristic analysis revealed superior performance of the double 10-fold cross-validation model for predicting sonic hedgehog, group 3, and group 4 tumors when using combined T1- and T2-weighted images (area under the curve = 0.79, 0.70, and 0.83, respectively). With the independent 3-dataset cross-validation strategy, select radiomic features were predictive of sonic hedgehog (area under the curve = 0.70-0.73) and group 4 (area under the curve = 0.76-0.80) medulloblastoma.

CONCLUSIONS

This study provides proof-of-concept results for the application of radiomic and machine learning approaches to a multi-institutional dataset for the prediction of medulloblastoma subgroups.

Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists

BACKGROUND

Chest radiograph interpretation is critical for the detection of thoracic diseases, including tuberculosis and lung cancer, which affect millions of people worldwide each year. This time-consuming task typically requires expert radiologists to read the images, leading to fatigue-based diagnostic error and lack of diagnostic expertise in areas of the world where radiologists are not available. Recently, deep learning approaches have been able to achieve expert-level performance in medical image interpretation tasks, powered by large network architectures and fueled by the emergence of large labeled datasets. The purpose of this study is to investigate the performance of a deep learning algorithm on the detection of pathologies in chest radiographs compared with practicing radiologists.

METHODS AND FINDINGS

We developed CheXNeXt, a convolutional neural network to concurrently detect the presence of 14 different pathologies, including pneumonia, pleural effusion, pulmonary masses, and nodules in frontal-view chest radiographs. CheXNeXt was trained and internally validated on the ChestX-ray8 dataset, with a held-out validation set consisting of 420 images, sampled to contain at least 50 cases of each of the original pathology labels. On this validation set, the majority vote of a panel of 3 board-certified cardiothoracic specialist radiologists served as reference standard. We compared CheXNeXt's discriminative performance on the validation set to the performance of 9 radiologists using the area under the receiver operating characteristic curve (AUC). The radiologists included 6 board-certified radiologists (average experience 12 years, range 4-28 years) and 3 senior radiology residents, from 3 academic institutions. We found that CheXNeXt achieved radiologist-level performance on 11 pathologies and did not achieve radiologist-level performance on 3 pathologies. The radiologists achieved statistically significantly higher AUC performance on cardiomegaly, emphysema, and hiatal hernia, with AUCs of 0.888 (95% confidence interval [CI] 0.863-0.910), 0.911 (95% CI 0.866-0.947), and 0.985 (95% CI 0.974-0.991), respectively, whereas CheXNeXt's AUCs were 0.831 (95% CI 0.790-0.870), 0.704 (95% CI 0.567-0.833), and 0.851 (95% CI 0.785-0.909), respectively. CheXNeXt performed better than radiologists in detecting atelectasis, with an AUC of 0.862 (95% CI 0.825-0.895), statistically significantly higher than radiologists' AUC of 0.808 (95% CI 0.777-0.838); there were no statistically significant differences in AUCs for the other 10 pathologies. The average time to interpret the 420 images in the validation set was substantially longer for the radiologists (240 minutes) than for CheXNeXt (1.5 minutes). The main limitations of our study are that neither CheXNeXt nor the radiologists were permitted to use patient history or review prior examinations and that evaluation was limited to a dataset from a single institution.

CONCLUSIONS

In this study, we developed and validated a deep learning algorithm that classified clinically important abnormalities in chest radiographs at a performance level comparable to practicing radiologists. Once tested prospectively in clinical settings, the algorithm could have the potential to expand patient access to chest radiograph diagnostics.

Deep-learning-assisted diagnosis for knee magnetic resonance imaging: Development and retrospective validation of MRNet

BACKGROUND

Magnetic 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 FINDINGS

Our 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 the panel of clinical experts.

CONCLUSIONS

Our deep learning model can rapidly generate accurate clinical pathology classifications of knee MRI exams from both internal and external datasets. Moreover, our results support the assertion that deep learning models can improve the performance of clinical experts during medical imaging interpretation. Further research is needed to validate the model prospectively and to determine its utility in the clinical setting.

Sex Differences in Cognitive Decline in Subjects with High Likelihood of Mild Cognitive Impairment due to Alzheimer's disease

Sex differences in Alzheimer’s disease (AD) biology and progression are not yet fully characterized. The goal of this study is to examine the effect of sex on cognitive progression in subjects with high likelihood of mild cognitive impairment (MCI) due to Alzheimer’s and followed up to 10 years in the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Cerebrospinal fluid total-tau and amyloid-beta (Aβ42) ratio values were used to sub-classify 559 MCI subjects (216 females, 343 males) as having “high” or “low” likelihood for MCI due to Alzheimer’s. Data were analyzed using mixed-effects models incorporating all follow-ups. The worsening from baseline in Alzheimer’s Disease Assessment Scale-Cognitive score (mean, SD) (9 ± 12) in subjects with high likelihood of MCI due to Alzheimer’s was markedly greater than that in subjects with low likelihood (1 ± 6, p < 0.0001). Among MCI due to AD subjects, the mean worsening in cognitive score was significantly greater in females (11.58 ± 14) than in males (6.87 ± 11, p = 0.006). Our findings highlight the need to further investigate these findings in other populations and develop sex specific timelines for Alzheimer’s disease progression.