Validation and estimation of spleen volume via computer-assisted segmentation on clinically acquired CT scans

Purpose: Deep learning is a promising technique for spleen segmentation. Our study aims to validate the reproducibility of deep learning-based spleen volume estimation by performing spleen segmentation on clinically acquired computed tomography (CT) scans from patients with myeloproliferative neoplasms. Approach: As approved by the institutional review board, we obtained 138 de-identified abdominal CT scans. A sum of voxel volume on an expert annotator's segmentations establishes the ground truth (estimation 1). We used our deep convolutional neural network (estimation 2) alongside traditional linear estimations (estimation 3 and 4) to estimate spleen volumes independently. Dice coefficient, Hausdorff distance,R 2 coefficient, Pearson R coefficient, the absolute difference in volume, and the relative difference in volume were calculated for 2 to 4 against the ground truth to compare and assess methods' performances. We re-labeled on scan-rescan on a subset of 40 studies to evaluate method reproducibility. Results: Calculated against the ground truth, theR 2 coefficients for our method (estimation 2) and linear method (estimation 3 and 4) are 0.998, 0.954, and 0.973, respectively. The Pearson R coefficients for the estimations against the ground truth are 0.999, 0.963, and 0.978, respectively (paired t -tests produced p < 0.05 between 2 and 3, and 2 and 4). Conclusion: The deep convolutional neural network algorithm shows excellent potential in rendering more precise spleen volume estimations. Our computer-aided segmentation exhibits reasonable improvements in splenic volume estimation accuracy.

High-resolution 3D abdominal segmentation with random patch network fusion

Deep learning for three dimensional (3D) abdominal organ segmentation on high-resolution computed tomography (CT) is a challenging topic, in part due to the limited memory provide by graphics processing units (GPU) and large number of parameters and in 3D fully convolutional networks (FCN). Two prevalent strategies, lower resolution with wider field of view and higher resolution with limited field of view, have been explored but have been presented with varying degrees of success. In this paper, we propose a novel patch-based network with random spatial initialization and statistical fusion on overlapping regions of interest (ROIs). We evaluate the proposed approach using three datasets consisting of 260 subjects with varying numbers of manual labels. Compared with the canonical "coarse-to-fine" baseline methods, the proposed method increases the performance on multi-organ segmentation from 0.799 to 0.856 in terms of mean DSC score (p-value < 0.01 with paired t-test). The effect of different numbers of patches is evaluated by increasing the depth of coverage (expected number of patches evaluated per voxel). In addition, our method outperforms other state-of-the-art methods in abdominal organ segmentation. In conclusion, the approach provides a memory-conservative framework to enable 3D segmentation on high-resolution CT. The approach is compatible with many base network structures, without substantially increasing the complexity during inference. Given a CT scan with at high resolution, a low-res section (left panel) is trained with multi-channel segmentation. The low-res part contains down-sampling and normalization in order to preserve the complete spatial information. Interpolation and random patch sampling (mid panel) is employed to collect patches. The high-dimensional probability maps are acquired (right panel) from integration of all patches on field of views.

BIAS: Transparent reporting of biomedical image analysis challenges

The number of biomedical image analysis challenges organized per year is steadily increasing. These international competitions have the purpose of benchmarking algorithms on common data sets, typically to identify the best method for a given problem. Recent research, however, revealed that common practice related to challenge reporting does not allow for adequate interpretation and reproducibility of results. To address the discrepancy between the impact of challenges and the quality (control), the Biomedical Image Analysis ChallengeS (BIAS) initiative developed a set of recommendations for the reporting of challenges. The BIAS statement aims to improve the transparency of the reporting of a biomedical image analysis challenge regardless of field of application, image modality or task category assessed. This article describes how the BIAS statement was developed and presents a checklist which authors of biomedical image analysis challenges are encouraged to include in their submission when giving a paper on a challenge into review. The purpose of the checklist is to standardize and facilitate the review process and raise interpretability and reproducibility of challenge results by making relevant information explicit.

Medial temporal lobe volumes in late-life depression: effects of age and vascular risk factors

Substantial work associates late-life depression with hippocampal pathology. However, there is less information about differences in hippocampal subfields and other connected temporal lobe regions and how these regions may be influenced by vascular factors. Individuals aged 60 years or older with and without a DSM-IV diagnosis of Major Depressive Disorder completed clinical assessments and 3 T cranial MRI using a protocol allowing for automated measurement of medial temporal lobe subfield volumes. A subset also completed pseudo-continuous arterial spin labeling, allowing for the measurement of hippocampal cerebral blood flow. In 59 depressed and 21 never-depressed elders (mean age = 66.4 years, SD = 5.8y, range 60-86y), the depressed group did not exhibit statistically significant volumetric differences for the total hippocampus or hippocampal subfields but did exhibit significantly smaller volumes of the perirhinal cortex, specifically in the BA36 region. Additionally, age had a greater effect in the depressed group on volumes of the cornu ammonis, entorhinal cortex, and BA36 region. Finally, both clinical and radiological markers of vascular risk were associated with smaller BA36 volumes, while reduced hippocampal blood flow was associated with smaller hippocampal and cornu ammonis volumes. In conclusion, while we did not observe group differences in hippocampal regions, we observed group differences and an effect of vascular pathology on the BA36 region, part of the perirhinal cortex. This is a critical region exhibiting atrophy in prodromal Alzheimer's disease. Moreover, the observed greater effect of age in the depressed groups is concordant with past longitudinal studies reporting greater hippocampal atrophy in late-life depression.

Microstructural Neuroimaging of Frailty in Cognitively Normal Older Adults

Physical frailty is an age-related clinical syndrome that is associated with multiple adverse health outcomes, including cognitive impairment and dementia. Recent studies have shown that frailty is associated with specific volumetric neuroimaging characteristics. Whether brain microstructural characteristics, particularly gray matter, associated with frailty exist and what their spatial distribution is have not been explored. We identified 670 participants of the Baltimore Longitudinal Study of Aging who were aged 60 and older and cognitively normal and who had concurrent data on frailty and regional microstructural neuroimaging markers by diffusion tensor imaging (DTI), including mean diffusivity (MD) of gray matter and fractional anisotropy (FA) of white matter. We identified neuroimaging markers that were associated with frailty status (non-frail, pre-frail, frail) and further examined differences between three groups using multivariate linear regression (non-frail = reference). Models were adjusted for age, sex, race, years of education, body mass index, scanner type, and Apolipoprotein E e4 carrier status. Compared to the non-frail participants, those who were frail had higher MD in the medial frontal cortex, several subcortical regions (putamen, caudate, thalamus), anterior cingulate cortex, and a trend of lower FA in the body of the corpus callosum. Those who were pre-frail also had higher MD in the putamen and a trend of lower FA in the body of the corpus callosum. Our study demonstrates for the first time that the microstructure of both gray and white matter differs by frailty status in cognitively normal older adults. Brain areas were not widespread but mostly localized in frontal and subcortical motor areas and the body of the corpus callosum. Whether changes in brain microstructure precede future frailty development warrants further investigation.

Generalizing deep whole-brain segmentation for post-contrast MRI with transfer learning

Purpose: Generalizability is an important problem in deep neural networks, especially with variability of data acquisition in clinical magnetic resonance imaging (MRI). Recently, the spatially localized atlas network tiles (SLANT) can effectively segment whole brain, non-contrast T1w MRI with 132 volumetric labels. Transfer learning (TL) is a commonly used domain adaptation tool to update the neural network weights for local factors, yet risks degradation of performance on the original validation/test cohorts. Approach: We explore TL using unlabeled clinical data to address these concerns in the context of adapting SLANT to scanning protocol variations. We optimize whole-brain segmentation on heterogeneous clinical data by leveraging 480 unlabeled pairs of clinically acquired T1w MRI with and without intravenous contrast. We use labels generated on the pre-contrast image to train on the post-contrast image in a five-fold cross-validation framework. We further validated on a withheld test set of 29 paired scans over a different acquisition domain. Results: Using TL, we improve reproducibility across imaging pairs measured by the reproducibility Dice coefficient (rDSC) between the pre- and post-contrast image. We showed an increase over the original SLANT algorithm (rDSC 0.82 versus 0.72) and the FreeSurfer v6.0.1 segmentation pipeline ( rDSC = 0.53 ). We demonstrate the impact of this work decreasing the root-mean-squared error of volumetric estimates of the hippocampus between paired images of the same subject by 67%. Conclusion: This work demonstrates a pipeline for unlabeled clinical data to translate algorithms optimized for research data to generalize toward heterogeneous clinical acquisitions.

Lower cardiac output is associated with neurodegeneration among older adults with normal cognition but not mild cognitive impairment

Subclinical cardiac dysfunction is associated with smaller total brain volume on magnetic resonance imaging (MRI). To study whether cardiac output relates to regional measurements of grey and white matter structure, older adults (n = 326) underwent echocardiogram to quantify cardiac output (L/min) and brain MRI. Linear regressions related cardiac output to grey matter volumes measured on T₁ and white matter hyperintensities assessed on T₂-FLAIR. Voxelwise analyses related cardiac output to diffusion tensor imaging adjusting for demographic, genetic, and vascular risk factors. Follow-up models assessed a cardiac output x diagnosis interaction with stratification (normal cognition, mild cognitive impairment). Cardiac output interacted with diagnosis, such that lower cardiac output related to smaller total grey matter (p = 0.01), frontal lobe (p = 0.01), and occipital lobe volumes (p = 0.01) among participants with normal cognition. When excluding participants with cardiovascular disease and atrial fibrillation, associations emerged with smaller parietal lobe (p = 0.005) and hippocampal volume (p = 0.05). Subtle age-related cardiac changes may disrupt neuronal homeostasis and impact grey matter integrity prior to cognitive impairment.

Semi-supervised Machine Learning with MixMatch and Equivalence Classes

Semi-supervised methods have an increasing impact on computer vision tasks to make use of scarce labels on large datasets, yet these approaches have not been well translated to medical imaging. Of particular interest, the MixMatch method achieves significant performance improvement over popular semi-supervised learning methods with scarce labels in the CIFAR-10 dataset. In a complementary approach, Nullspace Tuning on equivalence classes offers the potential to leverage multiple subject scans when the ground truth for the subject is unknown. This work is the first to (1) explore MixMatch with Nullspace Tuning in the context of medical imaging and (2) characterize the impacts of the methods with diminishing labels. We consider two distinct medical imaging domains: skin lesion diagnosis and lung cancer prediction. In both cases we evaluate models trained with diminishing labeled data using supervised, MixMatch, and Nullspace Tuning methods as well as MixMatch with Nullspace Tuning together. MixMatch with Nullspace Tuning together is able to achieve an AUC of 0.755 in lung cancer diagnosis with only 200 labeled subjects on the National Lung Screening Trial and a balanced multi-class accuracy of 77% with only 779 labeled examples on HAM10000. This performance is similar to that of the fully supervised methods when all labels are available. In advancing data driven methods in medical imaging, it is important to consider the use of current state-of-the-art semi-supervised learning methods from the greater machine learning community and their impact on the limitations of data acquisition and annotation.

Federated Gradient Averaging for Multi-Site Training with Momentum-Based Optimizers

Multi-site training methods for artificial neural networks are of particular interest to the medical machine learning community primarily due to the difficulty of data sharing between institutions. However, contemporary multi-site techniques such as weight averaging and cyclic weight transfer make theoretical sacrifices to simplify implementation. In this paper, we implement federated gradient averaging (FGA), a variant of federated learning without data transfer that is mathematically equivalent to single site training with centralized data. We evaluate two scenarios: a simulated multi-site dataset for handwritten digit classification with MNIST and a real multi-site dataset with head CT hemorrhage segmentation. We compare federated gradient averaging to single site training, federated weight averaging (FWA), and cyclic weight transfer. In the MNIST task, we show that training with FGA results in a weight set equivalent to centralized single site training. In the hemorrhage segmentation task, we show that FGA achieves on average superior results to both FWA and cyclic weight transfer due to its ability to leverage momentum-based optimization.

Prediction of Type II Diabetes Onset with Computed Tomography and Electronic Medical Records

Type II diabetes mellitus (T2DM) is a significant public health concern with multiple known risk factors (e.g., body mass index (BMI), body fat distribution, glucose levels). Improved prediction or prognosis would enable earlier intervention before possibly irreversible damage has occurred. Meanwhile, abdominal computed tomography (CT) is a relatively common imaging technique. Herein, we explore secondary use of the CT imaging data to refine the risk profile of future diagnosis of T2DM. In this work, we delineate quantitative information and imaging slices of patient history to predict onset T2DM retrieved from ICD-9 codes at least one year in the future. Furthermore, we investigate the role of five different types of electronic medical records (EMR), specifically 1) demographics; 2) pancreas volume; 3) visceral/subcutaneous fat volumes in L2 region of interest; 4) abdominal body fat distribution and 5) glucose lab tests in prediction. Next, we build a deep neural network to predict onset T2DM with pancreas imaging slices. Finally, motivated by multi-modal machine learning, we construct a merged framework to combine CT imaging slices with EMR information to refine the prediction. We empirically demonstrate our proposed joint analysis involving images and EMR leads to 4.25% and 6.93% AUC increase in predicting T2DM compared with only using images or EMR. In this study, we used case-control dataset of 997 subjects with CT scans and contextual EMR scores. To the best of our knowledge, this is the first work to show the ability to prognose T2DM using the patients' contextual and imaging history. We believe this study has promising potential for heterogeneous data analysis and multi-modal medical applications.

Time-distanced gates in long short-term memory networks

The Long Short-Term Memory (LSTM) network is widely used in modeling sequential observations in fields ranging from natural language processing to medical imaging. The LSTM has shown promise for interpreting computed tomography (CT) in lung screening protocols. Yet, traditional image-based LSTM models ignore interval differences, while recently proposed interval-modeled LSTM variants are limited in their ability to interpret temporal proximity. Meanwhile, clinical imaging acquisition may be irregularly sampled, and such sampling patterns may be commingled with clinical usages. In this paper, we propose the Distanced LSTM (DLSTM) by introducing time-distanced (i.e., time distance to the last scan) gates with a temporal emphasis model (TEM) targeting at lung cancer diagnosis (i.e., evaluating the malignancy of pulmonary nodules). Briefly, (1) the time distance of every scan to the last scan is modeled explicitly, (2) time-distanced input and forget gates in DLSTM are introduced across regular and irregular sampling sequences, and (3) the newer scan in serial data is emphasized by the TEM. The DLSTM algorithm is evaluated with both simulated data and real CT images (from 1794 National Lung Screening Trial (NLST) patients with longitudinal scans and 1420 clinical studied patients). Experimental results on simulated data indicate the DLSTM can capture families of temporal relationships that cannot be detected with traditional LSTM. Cross-validation on empirical CT datasets demonstrates that DLSTM achieves leading performance on both regularly and irregularly sampled data (e.g., improving LSTM from 0.6785 to 0.7085 on F1 score in NLST). In external-validation on irregularly acquired data, the benchmarks achieved 0.8350 (CNN feature) and 0.8380 (with LSTM) on AUC score, while the proposed DLSTM achieves 0.8905. In conclusion, the DLSTM approach is shown to be compatible with families of linear, quadratic, exponential, and log-exponential temporal models. The DLSTM can be readily extended with other temporal dependence interactions while hardly increasing overall model complexity.

CircleNet: Anchor-free Glomerulus Detection with Circle Representation

Object detection networks are powerful in computer vision, but not necessarily optimized for biomedical object detection. In this work, we propose CircleNet, a simple anchor-free detection method with circle representation for detection of the ball-shaped glomerulus. Different from the traditional bounding box based detection method, the bounding circle (1) reduces the degrees of freedom of detection representation, (2) is naturally rotation invariant, (3) and optimized for ball-shaped objects. The key innovation to enable this representation is the anchor-free framework with the circle detection head. We evaluate CircleNet in the context of detection of glomerulus. CircleNet increases average precision of the glomerulus detection from 0.598 to 0.647. Another key advantage is that CircleNet achieves better rotation consistency compared with bounding box representations.

Thalamocortical Anatomical Connectivity in Schizophrenia and Psychotic Bipolar Disorder

BACKGROUND

Anatomical connectivity between the thalamus and cortex, including the prefrontal cortex (PFC), is abnormal in schizophrenia. Overlapping phenotypes, including deficits in executive cognitive abilities dependent on PFC-thalamic circuitry, suggest dysrupted thalamocortical anatomical connectivity may extend to psychotic bipolar disorder. We tested this hypothesis and examined the impact of illness stage to inform when in the illness course thalamocortical dysconnectivity emerges.

METHODS

Diffusion-weighted imaging data were collected on 70 healthy individuals and 124 people with a psychotic disorder (schizophrenia spectrum = 75; psychotic bipolar disorder = 49), including 62 individuals in the early stage of psychosis. Anatomical connectivity between major divisions of the cortex and thalamus was quantified using probabilistic tractography and compared between groups. Associations between PFC-thalamic anatomical connectivity and executive cognitive abilities were examined using regression analysis.

RESULTS

Psychosis was associated with lower PFC-thalamic and elevated somatosensory-thalamic anatomical connectivity. Follow-up analyses established that lower PFC-thalamic and elevated somatosensory-thalamic anatomical connectivity were present in both schizophrenia and psychotic bipolar disorder. Lower PFC-thalamic anatomical connectivity was also present in early-stage and chronic psychosis. Contrary to expectations, lower PFC-thalamic anatomical connectivity was not associated with impaired executive cognitive abilities.

CONCLUSIONS

Altered thalamocortical anatomical connectivity, especially reduced PFC-thalamic connectivity, is a transdiagnostic feature of psychosis detectable in the early stage of illness. Further work is required to elucidate the functional consequences of the full spectrum of thalamocortical connectivity abnormalities in psychosis.

The future of digital health with federated learning

Data-driven machine learning (ML) has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare systems. Existing medical data is not fully exploited by ML primarily because it sits in data silos and privacy concerns restrict access to this data. However, without access to sufficient data, ML will be prevented from reaching its full potential and, ultimately, from making the transition from research to clinical practice. This paper considers key factors contributing to this issue, explores how federated learning (FL) may provide a solution for the future of digital health and highlights the challenges and considerations that need to be addressed.

Brainstem Functional Connectivity Disturbances in Epilepsy may Recover After Successful Surgery

BACKGROUND

Focal seizures in temporal lobe epilepsy (TLE) are associated with widespread brain network perturbations and neurocognitive problems.

OBJECTIVE

To determine whether brainstem connectivity disturbances improve with successful epilepsy surgery, as recent work has demonstrated decreased brainstem connectivity in TLE that is related to disease severity and neurocognitive profile.

METHODS

We evaluated 15 adult TLE patients before and after (>1 yr; mean, 3.4 yr) surgery, and 15 matched control subjects using magnetic resonance imaging to measure functional and structural connectivity of ascending reticular activating system (ARAS) structures, including cuneiform/subcuneiform nuclei (CSC), pedunculopontine nucleus (PPN), and ventral tegmental area (VTA).

RESULTS

TLE patients who achieved long-term postoperative seizure freedom (10 of 15) demonstrated increases in functional connectivity between ARAS structures and fronto-parietal-insular neocortex compared to preoperative baseline (P = .01, Kruskal-Wallis), with postoperative connectivity patterns resembling controls' connectivity. No functional connectivity changes were detected in 5 patients with persistent seizures after surgery (P = .9, Kruskal-Wallis). Among seizure-free postoperative patients, larger increases in CSC, PPN, and VTA functional connectivity were observed in individuals with more frequent seizures before surgery (P < .05 for each, Spearman's rho). Larger postoperative increases in PPN functional connectivity were seen in patients with lower baseline verbal IQ (P = .03, Spearman's rho) or verbal memory (P = .04, Mann-Whitney U). No changes in ARAS structural connectivity were detected after successful surgery.

CONCLUSION

ARAS functional connectivity disturbances are present in TLE but may recover after successful epilepsy surgery. Larger increases in postoperative connectivity may be seen in individuals with more severe disease at baseline.

Distortion correction of diffusion weighted MRI without reverse phase-encoding scans or field-maps

Diffusion magnetic resonance images may suffer from geometric distortions due to susceptibility induced off resonance fields, which cause geometric mismatch with anatomical images and ultimately affect subsequent quantification of microstructural or connectivity indices. State-of-the art diffusion distortion correction methods typically require data acquired with reverse phase encoding directions, resulting in varying magnitudes and orientations of distortion, which allow estimation of an undistorted volume. Alternatively, additional field maps acquisitions can be used along with sequence information to determine warping fields. However, not all imaging protocols include these additional scans and cannot take advantage of state-of-the art distortion correction. To avoid additional acquisitions, structural MRI (undistorted scans) can be used as registration targets for intensity driven correction. In this study, we aim to (1) enable susceptibility distortion correction with historical and/or limited diffusion datasets that do not include specific sequences for distortion correction and (2) avoid the computationally intensive registration procedure typically required for distortion correction using structural scans. To achieve these aims, we use deep learning (3D U-nets) to synthesize an undistorted b0 image that matches geometry of structural T1w images and intensity contrasts from diffusion images. Importantly, the training dataset is heterogenous, consisting of varying acquisitions of both structural and diffusion. We apply our approach to a withheld test set and show that distortions are successfully corrected after processing. We quantitatively evaluate the proposed distortion correction and intensity-based registration against state-of-the-art distortion correction (FSL topup). The results illustrate that the proposed pipeline results in b0 images that are geometrically similar to non-distorted structural images, and more closely match state-of-the-art correction with additional acquisitions. In addition, we show generalizability of the proposed approach to datasets that were not in the original training / validation / testing datasets. These datasets included varying populations, contrasts, resolutions, and magnitudes and orientations of distortion and show efficacious distortion correction. The method is available as a Singularity container, source code, and an executable trained model to facilitate evaluation.

A bayesian approach to examining default mode network functional connectivity and cognitive performance in major depressive disorder

To reconcile the inconsistency of the association between the resting-state functional connectivity (RSFC) and cognitive performance in healthy and depressed groups due to high variance of both measures, we proposed a Bayesian spatio-temporal model to precisely and accurately estimate the RSFC in depressed and nondepressed participants. This model was employed to estimate spatially-adjusted functional connectivity (saFC) in the extended default mode network (DMN) that was hypothesized to correlate with cognitive performance in both depressed and nondepressed. Multiple linear regression models were used to study the relationship between DMN saFC and cognitive performance scores measured in the following four cognitive domains while adjusting for age, sex, and education. In ROI pairs including the posterior cingulate (PCC) and anterior cingulate (ACC) cortex regions, the relationship between connectivity and cognition was found only with the Bayesian approach. Moreover, only the Bayesian approach was able to detect a significant diagnostic difference in the association in ROI pairs, including both PCC and ACC regions, due to smaller variance for the saFC estimator. The results confirm that a reliable and precise saFC estimator, based on the Bayesian model, can foster scientific discovery that may not be feasible with the conventional ROI-based FC estimator (denoted as 'AVG-FC').

Lasting consequences of concussion on the aging brain: Findings from the Baltimore Longitudinal Study of Aging

Studies suggest that concussions may be related to increased risk of neurodegenerative diseases, such as Chronic Traumatic Encephalopathy and Alzheimer’s Disease. Most neuroimaging studies show effects of concussionsin frontal and temporal lobes of the brain, yet the long-term impacts of concussions on the aging brain have not been well studied. We examined neuroimaging data from 51 participants (mean age at first imaging visit = 65.1±11.23) in the Baltimore Longitudinal Study of Aging (BLSA) who reported a concussion in their medical history an average of 23 years prior to the first imaging visit, and compared them to 150 participants (mean age at first imaging visit = 66.6 ± 10.97) with no history of concussion. Participants underwent serial structural MRI overa mean of 5.17 ± 6.14 years and DTI over a mean of 2.92 ± 2.22 years to measure brain structure, as well as ¹⁵O-water PET over a mean of 5.33 ± 2.19 years to measure brain function. A battery of neuropsychological tests was also administered over a mean of 11.62 ± 7.41 years. Analyses of frontal and temporal lobe regions were performed to examine differences in these measures between the concussion and control groups at first imaging visit and in change over time. Compared to those without concussion, participants with a prior concussion had greater brain atrophy in temporal lobe white matter and hippocampus at first imaging visit, which remained stable throughout the follow-up visits. Those with prior concussion also showed differences in white matter microstructure using DTI, including increased radial and axial diffusivity in the fornix/stria terminalis, anterior corona radiata, and superior longitudinal fasciculus at first imaging visit. In ¹⁵O-water PET, higher resting cerebral blood flow was seen at first imaging visit in orbitofrontal and lateral temporal regions, and both increases and decreases were seen in prefrontal, cingulate, insular, hippocampal, and ventral temporal regions with longitudinal follow-up. There were no significant differences in neuropsychological performance between groups. Most of the differences observed between the concussed and non-concussed groups were seen at the first imaging visit, suggesting that concussions can produce long-lasting structural and functional alterations in temporal and frontal regions of the brain in older individuals. These results also suggest that many of the reported short-term effects of concussion may still be apparent later in life.

Assessing the Accuracy of a Deep Learning Method to Risk Stratify Indeterminate Pulmonary Nodules

Rationale: The management of indeterminate pulmonary nodules (IPNs) remains challenging, resulting in invasive procedures and delays in diagnosis and treatment. Strategies to decrease the rate of unnecessary invasive procedures and optimize surveillance regimens are needed.Objectives: To develop and validate a deep learning method to improve the management of IPNs.Methods: A Lung Cancer Prediction Convolutional Neural Network model was trained using computed tomography images of IPNs from the National Lung Screening Trial, internally validated, and externally tested on cohorts from two academic institutions.Measurements and Main Results: The areas under the receiver operating characteristic curve in the external validation cohorts were 83.5% (95% confidence interval [CI], 75.4-90.7%) and 91.9% (95% CI, 88.7-94.7%), compared with 78.1% (95% CI, 68.7-86.4%) and 81.9 (95% CI, 76.1-87.1%), respectively, for a commonly used clinical risk model for incidental nodules. Using 5% and 65% malignancy thresholds defining low- and high-risk categories, the overall net reclassifications in the validation cohorts for cancers and benign nodules compared with the Mayo model were 0.34 (Vanderbilt) and 0.30 (Oxford) as a rule-in test, and 0.33 (Vanderbilt) and 0.58 (Oxford) as a rule-out test. Compared with traditional risk prediction models, the Lung Cancer Prediction Convolutional Neural Network was associated with improved accuracy in predicting the likelihood of disease at each threshold of management and in our external validation cohorts.Conclusions: This study demonstrates that this deep learning algorithm can correctly reclassify IPNs into low- or high-risk categories in more than a third of cancers and benign nodules when compared with conventional risk models, potentially reducing the number of unnecessary invasive procedures and delays in diagnosis.

Multi-path x-D Recurrent Neural Networks for Collaborative Image Classification

With the rapid development of image acquisition and storage, multiple images per class are commonly available for computer vision tasks (e.g., face recognition, object detection, medical imaging, etc.). Recently, the recurrent neural network (RNN) has been widely integrated with convolutional neural networks (CNN) to perform image classification on ordered (sequential) data. In this paper, by permutating multiple images as multiple dummy orders, we generalize the ordered "RNN+CNN" design (longitudinal) to a novel unordered fashion, called Multi-path x-D Recurrent Neural Network (MxDRNN) for image classification. To the best of our knowledge, few (if any) existing studies have deployed the RNN framework to unordered intra-class images to leverage classification performance. Specifically, multiple learning paths are introduced in the MxDRNN to extract discriminative features by permutating input dummy orders. Eight datasets from five different fields (MNIST, 3D-MNIST, CIFAR, VGGFace2, and lung screening computed tomography) are included to evaluate the performance of our method. The proposed MxDRNN improves the baseline performance by a large margin across the different application fields (e.g., accuracy from 46.40% to 76.54% in VGGFace2 test pose set, AUC from 0.7418 to 0.8162 in NLST lung dataset). Additionally, empirical experiments show the MxDRNN is more robust to category-irrelevant attributes (e.g., expression, pose in face images), which may introduce difficulties for image classification and algorithm generalizability. The code is publicly available.

Using phecode analysis to characterize co-occurring medical conditions in autism spectrum disorder

LAY ABSTRACT

People with autism spectrum disorder often have a number of other medical conditions in addition to autism. These can range from constipation to epilepsy. This study uses medical record data to understand how frequently and how long people with autism have to be seen by a medical professional for these other medical conditions. This study confirmed that people with autism often have a number of other medical conditions and that they have to go see a medical professional about those conditions often. We also looked to see if children diagnosed with autism after age 5 years might have different medical conditions compared to children diagnosed earlier. Children diagnosed later had more conditions like asthma, hearing loss, and mood disorders. This work describes how much medical care people with autism get for different medical conditions and the burden of seeking additional medical care for people with autism and their families.

Cross-scanner and cross-protocol multi-shell diffusion MRI data harmonization: Algorithms and results

Cross-scanner and cross-protocol variability of diffusion magnetic resonance imaging (dMRI) data are known to be major obstacles in multi-site clinical studies since they limit the ability to aggregate dMRI data and derived measures. Computational algorithms that harmonize the data and minimize such variability are critical to reliably combine datasets acquired from different scanners and/or protocols, thus improving the statistical power and sensitivity of multi-site studies. Different computational approaches have been proposed to harmonize diffusion MRI data or remove scanner-specific differences. To date, these methods have mostly been developed for or evaluated on single b-value diffusion MRI data. In this work, we present the evaluation results of 19 algorithms that are developed to harmonize the cross-scanner and cross-protocol variability of multi-shell diffusion MRI using a benchmark database. The proposed algorithms rely on various signal representation approaches and computational tools, such as rotational invariant spherical harmonics, deep neural networks and hybrid biophysical and statistical approaches. The benchmark database consists of data acquired from the same subjects on two scanners with different maximum gradient strength (80 and 300 ​mT/m) and with two protocols. We evaluated the performance of these algorithms for mapping multi-shell diffusion MRI data across scanners and across protocols using several state-of-the-art imaging measures. The results show that data harmonization algorithms can reduce the cross-scanner and cross-protocol variabilities to a similar level as scan-rescan variability using the same scanner and protocol. In particular, the LinearRISH algorithm based on adaptive linear mapping of rotational invariant spherical harmonics features yields the lowest variability for our data in predicting the fractional anisotropy (FA), mean diffusivity (MD), mean kurtosis (MK) and the rotationally invariant spherical harmonic (RISH) features. But other algorithms, such as DIAMOND, SHResNet, DIQT, CMResNet show further improvement in harmonizing the return-to-origin probability (RTOP). The performance of different approaches provides useful guidelines on data harmonization in future multi-site studies.

Learning from dispersed manual annotations with an optimized data weighting policy

Purpose: Deep learning methods have become essential tools for quantitative interpretation of medical imaging data, but training these approaches is highly sensitive to biases and class imbalance in the available data. There is an opportunity to increase the available training data by combining across different data sources (e.g., distinct public projects); however, data collected under different scopes tend to have differences in class balance, label availability, and subject demographics. Recent work has shown that importance sampling can be used to guide training selection. To date, these approaches have not considered imbalanced data sources with distinct labeling protocols. Approach: We propose a sampling policy, known as adaptive stochastic policy (ASP), inspired by reinforcement learning to adapt training based on subject, data source, and dynamic use criteria. We apply ASP in the context of multiorgan abdominal computed tomography segmentation. Training was performed with cross validation on 840 subjects from 10 data sources. External validation was performed with 20 subjects from 1 data source. Results: Four alternative strategies were evaluated with the state-of-the-art baseline as upper confident bound (UCB). ASP achieves average Dice of 0.8261 compared to 0.8135 UCB ( p < 0.01 , paired t -test) across fivefold cross validation. On withheld testing datasets, the proposed ASP achieved 0.8265 mean Dice versus 0.8077 UCB ( p < 0.01 , paired t -test). Conclusions: ASP provides a flexible reweighting technique for training deep learning models. We conclude that the proposed method adapts the sample importance, which leverages the performance on a challenging multisite, multiorgan, and multisize segmentation task.