RESUMO
BACKGROUND: Alzheimer disease (AD) and other dementias are associated with vascular changes and amyloid deposition, which may be reflected as density changes in the retinal capillaries. These changes may can be directly visualized and quantified with optical coherence tomography angiography (OCTA), making OCTA a potential noninvasive preclinical biomarker of small vessel disease and amyloid positivity. Our objective was to investigate the feasibility of retinal imaging metrics as noninvasive biomarkers of small vessel disease and amyloid positivity in the brain. METHODS: We investigated associations between OCTA and neuroimaging and cognitive metrics in 41 participants without dementia from the Mayo Clinic Study of Aging and Alzheimer's Disease Research Center. OCTA metrics included superficial, deep, and full retina capillary density of the fovea, parafovea, and macula as well as the area of the foveal avascular zone (FAZ). Neuroimaging metrics included a high burden of white matter hyperintensity (WMH), presence of cerebral microbleeds (CMB), lacunar infarcts, and amyloid positivity as evidenced on positron emission tomography (PET), whereas cognitive metrics included mini-mental status examination (MMSE) score. We performed generalized estimating equations to account for measurements in each eye while controlling for age and sex to estimate associations between OCTA metrics and neuroimaging and cognitive scores. RESULTS: Associations between OCTA and neuroimaging metrics were restricted to the fovea. OCTA showed decreased capillary density with high burden of WMH in both the superficial (P = 0.003), deep (P = 0.004), and full retina (P = 0.01) in the fovea but not the parafovea or whole macula. Similarly, participants with amyloid PET positivity had significantly decreased capillary density in the superficial fovea (P = 0.027) and deep fovea (P = 0.03) but higher density in the superficial parafovea (P = 0.038). Participants with amyloid PET positivity also had a significantly larger FAZ (P = 0.031), whereas in those with high WMH burden the difference did not reach statistical significance (P = 0.075). There was also a positive association between MMSE and capillary density of the full retina within the fovea (P = 0.037) and in the superficial parafovea (P = 0.046). No associations were found between OCTA metrics and presence of CMB or presence of lacunar infarcts. CONCLUSION: The associations of lower foveal capillary density with cerebral WMH and amyloid positivity suggest that further research is warranted to evaluate for shared mechanisms of disease between small vessel disease and AD pathologies.
RESUMO
While plasma biomarkers for Alzheimer's disease (AD) are increasingly being evaluated for clinical diagnosis and prognosis, few population-based autopsy studies have evaluated their utility in the context of predicting neuropathological changes. Our goal was to investigate the utility of clinically available plasma markers in predicting Braak staging, neuritic plaque score, Thal phase, and overall AD neuropathological change (ADNC).We utilized a population-based prospective study of 350 participants with autopsy and antemortem plasma biomarker testing using clinically available antibody assay (Quanterix) consisting of Aß42/40 ratio, p-tau181, GFAP, and NfL. We utilized a variable selection procedure in cross-validated (CV) logistic regression models to identify the best set of plasma predictors along with demographic variables, and a subset of neuropsychological tests comprising the Mayo Clinic Preclinical Alzheimer Cognitive Composite (Mayo-PACC). ADNC was best predicted with plasma GFAP, NfL, p-tau181 biomarkers along with APOE ε4 carrier status and Mayo-PACC cognitive score (CV AUC = 0.798). Braak staging was best predicted using plasma GFAP, p-tau181, and cognitive scores (CV AUC = 0.774). Neuritic plaque score was best predicted using plasma Aß42/40 ratio, p-tau181, GFAP, and NfL biomarkers (CV AUC = 0.770). Thal phase was best predicted using GFAP, NfL, p-tau181, APOE ε4 carrier status and Mayo-PACC cognitive score (CV AUC = 0.754). We found that GFAP and p-tau provided non-overlapping information on both neuritic plaque and Braak stage scores whereas Aß42/40 and NfL were mainly useful for prediction of neuritic plaque scores. Separating participants by cognitive status improved predictive performance, particularly when plasma biomarkers were included. Plasma biomarkers can differentially inform about overall ADNC pathology, Braak staging, and neuritic plaque score when combined with demographics and cognitive variables and have significant utility for earlier detection of AD.
Assuntos
Doença de Alzheimer , Humanos , Doença de Alzheimer/patologia , Placa Amiloide/patologia , Estudos Prospectivos , Apolipoproteína E4 , Biomarcadores , Proteínas tau , Peptídeos beta-AmiloidesRESUMO
OBJECTIVE: Late-life depression (LLD) is characterized by accelerated biological aging. Accelerated brain aging, estimated from structural magnetic resonance imaging (sMRI) data by a machine learning algorithm, is associated with LLD diagnosis, poorer cognitive performance, and disability. We hypothesized that accelerated brain aging moderates the antidepressant response. DESIGN AND INTERVENTIONS: Following MRI, participants entered an 8-week randomized, controlled trial of escitalopram. Nonremitting participants then entered an open-label 8-week trial of bupropion. PARTICIPANTS: Ninety-five individuals with LLD. MEASUREMENTS: A machine learning algorithm estimated each participant's brain age from sMRI data. This was used to calculate the brain-age gap (BAG), or how estimated age differed from chronological age. Secondary sMRI measures of aging pathology included white matter hyperintensity (WMH) volumes and hippocampal volumes. Mixed models examined the relationship between sMRI measures and change in depression severity. Initial analyses tested for a moderating effect of MRI measures on change in depression severity with escitalopram. Subsequent analyses tested for the effect of MRI measures on change in depression severity over time across trials. RESULTS: In the blinded initial phase, BAG was not significantly associated with a differential response to escitalopram over time. BAG was also not associated with a change in depression severity over time across both arms in the blinded phase or in the subsequent open-label bupropion phase. We similarly did not observe effects of WMH volume or hippocampal volume on change in depression severity over time. CONCLUSION: sMRI markers of accelerated brain aging were not associated with treatment response in this sequential antidepressant trial.
Assuntos
Bupropiona , Depressão , Envelhecimento/psicologia , Antidepressivos/uso terapêutico , Encéfalo/diagnóstico por imagem , Encéfalo/patologia , Depressão/psicologia , Humanos , Imageamento por Ressonância Magnética/métodosRESUMO
Deep learning has achieved impressive performance across a variety of tasks, including medical image processing. However, recent research has shown that deep neural networks are susceptible to small adversarial perturbations in the image. We study the impact of such adversarial perturbations in medical image processing where the goal is to predict an individual's age based on a 3D MRI brain image. We consider two models: a conventional deep neural network, and a hybrid deep learning model which additionally uses features informed by anatomical context. We find that we can introduce significant errors in predicted age by adding imperceptible noise to an image, can accomplish this even for large batches of images using a single perturbation, and that the hybrid model is much more robust to adversarial perturbations than the conventional deep neural network. Our work highlights limitations of current deep learning techniques in clinical applications, and suggests a path forward.
RESUMO
Detailed whole brain segmentation is an essential quantitative technique in medical image analysis, which provides a non-invasive way of measuring brain regions from a clinical acquired structural magnetic resonance imaging (MRI). Recently, deep convolution neural network (CNN) has been applied to whole brain segmentation. However, restricted by current GPU memory, 2D based methods, downsampling based 3D CNN methods, and patch-based high-resolution 3D CNN methods have been the de facto standard solutions. 3D patch-based high resolution methods typically yield superior performance among CNN approaches on detailed whole brain segmentation (>100 labels), however, whose performance are still commonly inferior compared with state-of-the-art multi-atlas segmentation methods (MAS) due to the following challenges: (1) a single network is typically used to learn both spatial and contextual information for the patches, (2) limited manually traced whole brain volumes are available (typically less than 50) for training a network. In this work, we propose the spatially localized atlas network tiles (SLANT) method to distribute multiple independent 3D fully convolutional networks (FCN) for high-resolution whole brain segmentation. To address the first challenge, multiple spatially distributed networks were used in the SLANT method, in which each network learned contextual information for a fixed spatial location. To address the second challenge, auxiliary labels on 5111 initially unlabeled scans were created by multi-atlas segmentation for training. Since the method integrated multiple traditional medical image processing methods with deep learning, we developed a containerized pipeline to deploy the end-to-end solution. From the results, the proposed method achieved superior performance compared with multi-atlas segmentation methods, while reducing the computational time from >30â¯h to 15â¯min. The method has been made available in open source (https://github.com/MASILab/SLANTbrainSeg).
Assuntos
Encéfalo/anatomia & histologia , Aprendizado Profundo , Processamento de Imagem Assistida por Computador/métodos , Imageamento Tridimensional/métodos , Atlas como Assunto , Humanos , Imageamento por Ressonância Magnética/métodos , Neuroimagem/métodosRESUMO
Cutaneous leishmaniasis (CL) is a neglected tropical disease that requires novel tools for its understanding, diagnosis, and treatment follow-up. In the cases of other cutaneous pathologies, such as cancer or cutaneous ulcers due to diabetes, optical diffuse reflectance-based tools and methods are widely used for the investigation of those illnesses. These types of tools and methods offer the possibility to develop portable diagnosis and treatment follow-up systems. In this article, we propose the use of a three-layer diffuse reflectance model for the study of the formation of cutaneous ulcers caused by CL. The proposed model together with an inverse-modeling procedure were used in the evaluation of diffuse-reflectance spectral signatures acquired from cutaneous ulcers formed in the dorsal area of 21 golden hamsters inoculated with Leishmanisis braziliensis. As result, the quantification of the model's variables related to the main biological parameters of skin were obtained, such as: diameter and volumetric fraction of keratinocytes, collagen; volumetric fraction of hemoglobin, and oxygen saturation. Those parameters show statistically significant differences among the different stages of the CL ulcer formation. We found that these differences are coherent with histopathological manifestations reported in the literature for the main phases of CL formation.
Assuntos
Leishmaniose Cutânea/patologia , Úlcera Cutânea/patologia , Pele/química , Espectrofotometria/métodos , Animais , Colágeno/fisiologia , Cricetinae , Modelos Animais de Doenças , Processamento Eletrônico de Dados , Feminino , Hemoglobinas/química , Leishmaniose Cutânea/metabolismo , Masculino , Mesocricetus , Oxigênio/química , Pele/patologia , Úlcera Cutânea/metabolismo , Úlcera Cutânea/parasitologiaRESUMO
OBJECTIVE: Current epilepsy therapies directed at altering the function of neurotransmitter receptors or ion channels, or release of synaptic vesicles fail to prevent seizures in approximately 30% of patients. A better understanding of the molecular mechanism underlying epilepsy is needed to provide new therapeutic targets. The activity of cyclic AMP (cAMP) response element-binding protein (CREB), a major transcription factor promoting CRE-mediated transcription, increases following a prolonged seizure called status epilepticus. It is also increased in the seizure focus of patients with medically intractable focal epilepsy. Herein we explored the effect of acute suppression of CREB activity on status epilepticus and spontaneous seizures in a chronic epilepsy model. METHODS: Pilocarpine chemoconvulsant was used to induce status epilepticus. To suppress CREB activity, a transgenic mouse line expressing an inducible dominant negative mutant of CREB (CREB(IR) ) with a serine to alanine 133 substitution was used. Status epilepticus and spontaneous seizures of transgenic and wild-type mice were analyzed using video-electroencephalography (EEG) to assess the effect of CREB suppression on seizures. RESULTS: Our findings indicate that activation of CREB(IR) shortens the duration of status epilepticus. The frequency of spontaneous seizures decreased in mice with chronic epilepsy during CREB(IR) induction; however, the duration of the spontaneous seizures was unchanged. Of interest, we found significantly reduced levels of phospho-CREB Ser133 upon activation of CREB(IR) , supporting prior work suggesting that binding to the CRE site is important for CREB phosphorylation. SIGNIFICANCE: Our results suggest that CRE transcription supports seizure activity both during status epilepticus and in spontaneous seizures. Thus, blocking of CRE transcription is a novel target for the treatment of epilepsy.
Assuntos
Proteína de Ligação a CREB/antagonistas & inibidores , Convulsivantes/farmacologia , Pilocarpina/farmacologia , Convulsões/fisiopatologia , Estado Epiléptico/fisiopatologia , Animais , Química Encefálica , Proteína de Ligação a CREB/análise , Modelos Animais de Doenças , Feminino , Immunoblotting , Masculino , Camundongos , Camundongos Endogâmicos C3H , Camundongos Endogâmicos C57BL , Camundongos Transgênicos , Fosforilação , Reação em Cadeia da Polimerase em Tempo Real , Convulsões/induzido quimicamente , Estado Epiléptico/induzido quimicamenteRESUMO
We report a case highlighting key clinical, CSF, and imaging findings of recurrent pleomorphic xanthoastrocytoma with leptomeningeal spread.
RESUMO
Heart failure with preserved ejection fraction (HFpEF) is an important, emerging risk factor for dementia, but it is not clear whether HFpEF contributes to a specific pattern of neuroanatomical changes in dementia. A major challenge to studying this is the relative paucity of datasets of patients with dementia, with/without HFpEF, and relevant neuroimaging. We sought to demonstrate the feasibility of using modern data mining tools to create and analyze clinical imaging datasets and identify the neuroanatomical signature of HFpEF-associated dementia. We leveraged the bioinformatics tools at Vanderbilt University Medical Center to identify patients with a diagnosis of dementia with and without comorbid HFpEF using the electronic health record. We identified high resolution, clinically-acquired neuroimaging data on 30 dementia patients with HFpEF (age 76.9 ± 8.12 years, 61% female) as well as 301 age- and sex-matched patients with dementia but without HFpEF to serve as comparators (age 76.2 ± 8.52 years, 60% female). We used automated image processing pipelines to parcellate the brain into 132 structures and quantify their volume. We found six regions with significant atrophy associated with HFpEF: accumbens area, amygdala, posterior insula, anterior orbital gyrus, angular gyrus, and cerebellar white matter. There were no regions with atrophy inversely associated with HFpEF. Patients with dementia and HFpEF have a distinct neuroimaging signature compared to patients with dementia only. Five of the six regions identified in are in the temporo-parietal region of the brain. Future studies should investigate mechanisms of injury associated with cerebrovascular disease leading to subsequent brain atrophy.
Assuntos
Demência , Insuficiência Cardíaca , Humanos , Feminino , Idoso , Idoso de 80 Anos ou mais , Masculino , Insuficiência Cardíaca/diagnóstico por imagem , Volume Sistólico , Função Ventricular Esquerda , Imageamento por Ressonância Magnética , Neuroimagem , Encéfalo/diagnóstico por imagem , Atrofia , Demência/diagnóstico por imagemRESUMO
Along with the increasing availability of electronic medical record (EMR) data, phenome-wide association studies (PheWAS) and phenome-disease association studies (PheDAS) have become a prominent, first-line method of analysis for uncovering the secrets of EMR. Despite this recent growth, there is a lack of approachable software tools for conducting these analyses on large-scale EMR cohorts. In this article, we introduce pyPheWAS, an open-source python package for conducting PheDAS and related analyses. This toolkit includes 1) data preparation, such as cohort censoring and age-matching; 2) traditional PheDAS analysis of ICD-9 and ICD-10 billing codes; 3) PheDAS analysis applied to a novel EMR phenotype mapping: current procedural terminology (CPT) codes; and 4) novelty analysis of significant disease-phenotype associations found through PheDAS. The pyPheWAS toolkit is approachable and comprehensive, encapsulating data prep through result visualization all within a simple command-line interface. The toolkit is designed for the ever-growing scale of available EMR data, with the ability to analyze cohorts of 100,000 + patients in less than 2 h. Through a case study of Down Syndrome and other intellectual developmental disabilities, we demonstrate the ability of pyPheWAS to discover both known and potentially novel disease-phenotype associations across different experiment designs and disease groups. The software and user documentation are available in open source at https://github.com/MASILab/pyPheWAS .
Assuntos
Registros Eletrônicos de Saúde , Estudo de Associação Genômica Ampla , Estudo de Associação Genômica Ampla/métodos , Fenótipo , SoftwareRESUMO
Body part regression is a promising new technique that enables content navigation through self-supervised learning. Using this technique, the global quantitative spatial location for each axial view slice is obtained from computed tomography (CT). However, it is challenging to define a unified global coordinate system for body CT scans due to the large variabilities in image resolution, contrasts, sequences, and patient anatomy. Therefore, the widely used supervised learning approach cannot be easily deployed. To address these concerns, we propose an annotation-free method named blind-unsupervised-supervision network (BUSN). The contributions of the work are in four folds: (1) 1030 multi-center CT scans are used in developing BUSN without any manual annotation. (2) the proposed BUSN corrects the predictions from unsupervised learning and uses the corrected results as the new supervision; (3) to improve the consistency of predictions, we propose a novel neighbor message passing (NMP) scheme that is integrated with BUSN as a statistical learning based correction; and (4) we introduce a new pre-processing pipeline with inclusion of the BUSN, which is validated on 3D multi-organ segmentation. The proposed method is trained on 1,030 whole body CT scans (230,650 slices) from five datasets, as well as an independent external validation cohort with 100 scans. From the body part regression results, the proposed BUSN achieved significantly higher median R-squared score (=0.9089) than the state-of-the-art unsupervised method (=0.7153). When introducing BUSN as a preprocessing stage in volumetric segmentation, the proposed pre-processing pipeline using BUSN approach increases the total mean Dice score of the 3D abdominal multi-organ segmentation from 0.7991 to 0.8145.
Assuntos
Corpo Humano , Tomografia Computadorizada por Raios X , Humanos , Processamento de Imagem Assistida por ComputadorRESUMO
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.
Assuntos
Imageamento Tridimensional , Redes Neurais de Computação , Abdome/diagnóstico por imagem , Humanos , Processamento de Imagem Assistida por Computador , Tomografia Computadorizada por Raios XRESUMO
PURPOSE: Dynamic contrast-enhanced computed tomography (CT) is widely used to provide dynamic tissue contrast for diagnostic investigation and vascular identification. However, the phase information of contrast injection is typically recorded manually by technicians, which introduces missing or mislabeling. Hence, imaging-based contrast phase identification is appealing, but challenging, due to large variations among different contrast protocols, vascular dynamics, and metabolism, especially for clinically acquired CT scans. The purpose of this study is to perform imaging-based phase identification for dynamic abdominal CT using a proposed adversarial learning framework across five representative contrast phases. METHODS: A generative adversarial network (GAN) is proposed as a disentangled representation learning model. To explicitly model different contrast phases, a low dimensional common representation and a class specific code are fused in the hidden layer. Then, the low dimensional features are reconstructed following a discriminator and classifier. 36 350 slices of CT scans from 400 subjects are used to evaluate the proposed method with fivefold cross-validation with splits on subjects. Then, 2216 slices images from 20 independent subjects are employed as independent testing data, which are evaluated using multiclass normalized confusion matrix. RESULTS: The proposed network significantly improved correspondence (0.93) over VGG, ResNet50, StarGAN, and 3DSE with accuracy scores 0.59, 0.62, 0.72, and 0.90, respectively (P < 0.001 Stuart-Maxwell test for normalized multiclass confusion matrix). CONCLUSION: We show that adversarial learning for discriminator can be benefit for capturing contrast information among phases. The proposed discriminator from the disentangled network achieves promising results.
Assuntos
Processamento de Imagem Assistida por Computador , Tomografia Computadorizada por Raios X , Abdome , Humanos , Tomografia Computadorizada EspiralRESUMO
Generalizability is an important problem in deep neural networks, especially in the context of the variability of data acquisition in clinical magnetic resonance imaging (MRI). Recently, the Spatially Localized Atlas Network Tiles (SLANT) approach has been shown to effectively segment whole brain non-contrast T1w MRI with 132 volumetric labels. Enhancing generalizability of SLANT would enable broader application of volumetric assessment in multi-site studies. Transfer learning (TL) is commonly to update neural network weights for local factors; yet, it is commonly recognized to risk degradation of performance on the original validation/test cohorts. Here, we explore TL by data augmentation to address these concerns in the context of adapting SLANT to anatomical variation (e.g., adults versus children) and scanning protocol (e.g., non-contrast research T1w MRI versus contrast-enhanced clinical T1w MRI). We consider two datasets: First, 30 T1w MRI of young children with manually corrected volumetric labels, and accuracy of automated segmentation defined relative to the manually provided truth. Second, 36 paired datasets of pre- and post-contrast clinically acquired T1w MRI, and accuracy of the post-contrast segmentations assessed relative to the pre-contrast automated assessment. For both studies, we augment the original TL step of SLANT with either only the new data or with both original and new data. Over baseline SLANT, both approaches yielded significantly improved performance (pediatric: 0.89 vs. 0.82 DSC, p<0.001; contrast: 0.80 vs 0.76, p<0.001). The performance on the original test set decreased with the new-data only transfer learning approach, so data augmentation was superior to strict transfer learning.
RESUMO
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.
RESUMO
PURPOSE: As deep neural networks achieve more success in the wide field of computer vision, greater emphasis is being placed on the generalizations of these models for production deployment. With sufficiently large training datasets, models can typically avoid overfitting their data; however, for medical imaging it is often difficult to obtain enough data from a single site. Sharing data between institutions is also frequently nonviable or prohibited due to security measures and research compliance constraints, enforced to guard protected health information (PHI) and patient anonymity. METHODS: In this paper, we implement cyclic weight transfer with independent datasets from multiple geographically disparate sites without compromising PHI. We compare results between single-site learning (SSL) and multisite learning (MSL) models on testing data drawn from each of the training sites as well as two other institutions. RESULTS: The MSL model attains an average dice similarity coefficient (DSC) of 0.690 on the holdout institution datasets with a volume correlation of 0.914, respectively corresponding to a 7% and 5% statistically significant improvement over the average of both SSL models, which attained an average DSC of 0.646 and average correlation of 0.871. CONCLUSIONS: We show that a neural network can be efficiently trained on data from two physically remote sites without consolidating patient data to a single location. The resulting network improves model generalization and achieves higher average DSCs on external datasets than neural networks trained on data from a single source.
Assuntos
Aprendizado Profundo , Hemorragia/diagnóstico por imagem , Processamento de Imagem Assistida por Computador/métodos , Tomografia Computadorizada por Raios X , HumanosRESUMO
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.
RESUMO
Dynamic contrast enhanced computed tomography (CT) is an imaging technique that provides critical information on the relationship of vascular structure and dynamics in the context of underlying anatomy. A key challenge for image processing with contrast enhanced CT is that phase discrepancies are latent in different tissues due to contrast protocols, vascular dynamics, and metabolism variance. Previous studies with deep learning frameworks have been proposed for classifying contrast enhancement with networks inspired by computer vision. Here, we revisit the challenge in the context of whole abdomen contrast enhanced CTs. To capture and compensate for the complex contrast changes, we propose a novel discriminator in the form of a multi-domain disentangled representation learning network. The goal of this network is to learn an intermediate representation that separates contrast enhancement from anatomy and enables classification of images with varying contrast time. Briefly, our unpaired contrast disentangling GAN(CD-GAN) Discriminator follows the ResNet architecture to classify a CT scan from different enhancement phases. To evaluate the approach, we trained the enhancement phase classifier on 21060 slices from two clinical cohorts of 230 subjects. The scans were manually labeled with three independent enhancement phases (non-contrast, portal venous and delayed). Testing was performed on 9100 slices from 30 independent subjects who had been imaged with CT scans from all contrast phases. Performance was quantified in terms of the multi-class normalized confusion matrix. The proposed network significantly improved correspondence over baseline UNet, ResNet50 and StarGAN's performance of accuracy scores 0.54. 0.55, 0.62 and 0.91, respectively (p-value<0.0001 paired t-test for ResNet versus CD-GAN). The proposed discriminator from the disentangled network presents a promising technique that may allow deeper modeling of dynamic imaging against patient specific anatomies.
RESUMO
Multiple instance learning (MIL) is a supervised learning methodology that aims to allow models to learn instance class labels from bag class labels, where a bag is defined to contain multiple instances. MIL is gaining traction for learning from weak labels but has not been widely applied to 3D medical imaging. MIL is well-suited to clinical CT acquisitions since (1) the highly anisotropic voxels hinder application of traditional 3D networks and (2) patch-based networks have limited ability to learn whole volume labels. In this work, we apply MIL with a deep convolutional neural network to identify whether clinical CT head image volumes possess one or more large hemorrhages (> 20cm3), resulting in a learned 2D model without the need for 2D slice annotations. Individual image volumes are considered separate bags, and the slices in each volume are instances. Such a framework sets the stage for incorporating information obtained in clinical reports to help train a 2D segmentation approach. Within this context, we evaluate the data requirements to enable generalization of MIL by varying the amount of training data. Our results show that a training size of at least 400 patient image volumes was needed to achieve accurate per-slice hemorrhage detection. Over a five-fold cross-validation, the leading model, which made use of the maximum number of training volumes, had an average true positive rate of 98.10%, an average true negative rate of 99.36%, and an average precision of 0.9698. The models have been made available along with source code1 to enabled continued exploration and adaption of MIL in CT neuroimaging.
RESUMO
Depression is associated with markers of accelerated aging, but it is unclear how this relationship changes across the lifespan. We examined whether a brain-based measure of accelerated aging differed between depressed and never-depressed subjects across the adult lifespan and whether it was related to cognitive performance and disability. We applied a machine-learning approach that estimated brain age from structural MRI data in two depressed cohorts, respectively 170 midlife adults and 154 older adults enrolled in studies with common entry criteria. Both cohorts completed broad cognitive batteries and the older subgroup completed a disability assessment. The machine-learning model estimated brain age from MRI data, which was compared to chronological age to determine the brain-age gap (BAG; estimated age-chronological age). BAG did not differ between midlife depressed and nondepressed adults. Older depressed adults exhibited significantly higher BAG than nondepressed elders (Wald χ2 = 8.84, p = 0.0029), indicating a higher estimated brain age than chronological age. BAG was not associated with midlife cognitive performance. In the older cohort, higher BAG was associated with poorer episodic memory performance (Wald χ2 = 4.10, p = 0.0430) and, in the older depressed group alone, slower processing speed (Wald χ2 = 4.43, p = 0.0354). We also observed a statistical interaction where greater depressive symptom severity in context of higher BAG was associated with poorer executive function (Wald χ2 = 5.89, p = 0.0152) and working memory performance (Wald χ2 = 4.47, p = 0.0346). Increased BAG was associated with greater disability (Wald χ2 = 6.00, p = 0.0143). Unlike midlife depression, geriatric depression exhibits accelerated brain aging, which in turn is associated with cognitive and functional deficits.