Dataset Information

Convolutional Neural Networks for Fully Automated Diagnosis of Cardiac Amyloidosis by Cardiac Magnetic Resonance Imaging.

ABSTRACT:

Aims

We tested the hypothesis that artificial intelligence (AI)-powered algorithms applied to cardiac magnetic resonance (CMR) images could be able to detect the potential patterns of cardiac amyloidosis (CA). Readers in CMR centers with a low volume of referrals for the detection of myocardial storage diseases or a low volume of CMRs, in general, may overlook CA. In light of the growing prevalence of the disease and emerging therapeutic options, there is an urgent need to avoid misdiagnoses.

Methods and results

Using CMR data from 502 patients (CA: n = 82), we trained convolutional neural networks (CNNs) to automatically diagnose patients with CA. We compared the diagnostic accuracy of different state-of-the-art deep learning techniques on common CMR imaging protocols in detecting imaging patterns associated with CA. As a result of a 10-fold cross-validated evaluation, the best-performing fine-tuned CNN achieved an average ROC AUC score of 0.96, resulting in a diagnostic accuracy of 94% sensitivity and 90% specificity.

Conclusions

Applying AI to CMR to diagnose CA may set a remarkable milestone in an attempt to establish a fully computational diagnostic path for the diagnosis of CA, in order to support the complex diagnostic work-up requiring a profound knowledge of experts from different disciplines.

SUBMITTER: Agibetov A

PROVIDER: S-EPMC8705947 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Cardiovascular resonance (CMR) imaging is a standard imaging modality for assessing cardiovascular diseases (CVDs), the leading cause of death globally. CMR enables accurate quantification of the cardiac chamber volume, ejection fraction and myocardial mass, providing information for diagnosis and monitoring of CVDs. However, for years, clinicians have been relying on manual approaches for CMR image analysis, which is time consuming and prone to subjective errors. It is a major clinical challenge to automatically derive quantitative and clinically relevant information from CMR images.Deep neural networks have shown a great potential in image pattern recognition and segmentation for a variety of tasks. Here we demonstrate an automated analysis method for CMR images, which is based on a fully convolutional network (FCN). The network is trained and evaluated on a large-scale dataset from the UK Biobank, consisting of 4,875 subjects with 93,500 pixelwise annotated images. The performance of the method has been evaluated using a number of technical metrics, including the Dice metric, mean contour distance and Hausdorff distance, as well as clinically relevant measures, including left ventricle (LV) end-diastolic volume (LVEDV) and end-systolic volume (LVESV), LV mass (LVM); right ventricle (RV) end-diastolic volume (RVEDV) and end-systolic volume (RVESV).By combining FCN with a large-scale annotated dataset, the proposed automated method achieves a high performance in segmenting the LV and RV on short-axis CMR images and the left atrium (LA) and right atrium (RA) on long-axis CMR images. On a short-axis image test set of 600 subjects, it achieves an average Dice metric of 0.94 for the LV cavity, 0.88 for the LV myocardium and 0.90 for the RV cavity. The mean absolute difference between automated measurement and manual measurement is 6.1 mL for LVEDV, 5.3 mL for LVESV, 6.9 gram for LVM, 8.5 mL for RVEDV and 7.2 mL for RVESV. On long-axis image test sets, the average Dice metric is 0.93 for the LA cavity (2-chamber view), 0.95 for the LA cavity (4-chamber view) and 0.96 for the RA cavity (4-chamber view). The performance is comparable to human inter-observer variability.We show that an automated method achieves a performance on par with human experts in analysing CMR images and deriving clinically relevant measures.

Project description:BackgroundCardiovascular magnetic resonance (CMR) myocardial native T1 mapping allows assessment of interstitial diffuse fibrosis. In this technique, the global and regional T1 are measured manually by drawing region of interest in motion-corrected T1 maps. The manual analysis contributes to an already lengthy CMR analysis workflow and impacts measurements reproducibility. In this study, we propose an automated method for combined myocardium segmentation, alignment, and T1 calculation for myocardial T1 mapping.MethodsA deep fully convolutional neural network (FCN) was used for myocardium segmentation in T1 weighted images. The segmented myocardium was then resampled on a polar grid, whose origin is located at the center-of-mass of the segmented myocardium. Myocardium T1 maps were reconstructed from the resampled T1 weighted images using curve fitting. The FCN was trained and tested using manually segmented images for 210 patients (5 slices, 11 inversion times per patient). An additional image dataset for 455 patients (5 slices and 11 inversion times per patient), analyzed by an expert reader using a semi-automatic tool, was used to validate the automatically calculated global and regional T1 values. Bland-Altman analysis, Pearson correlation coefficient, r, and the Dice similarity coefficient (DSC) were used to evaluate the performance of the FCN-based analysis on per-patient and per-slice basis. Inter-observer variability was assessed using intraclass correlation coefficient (ICC) of the T1 values calculated by the FCN-based automatic method and two readers.ResultsThe FCN achieved fast segmentation (<?0.3?s/image) with high DSC (0.85?±?0.07). The automatically and manually calculated T1 values (1091?±?59?ms and 1089?±?59?ms, respectively) were highly correlated in per-patient (r?=?0.82; slope?=?1.01; p?<?0.0001) and per-slice (r?=?0.72; slope?=?1.01; p?<?0.0001) analyses. Bland-Altman analysis showed good agreement between the automated and manual measurements with 95% of measurements within the limits-of-agreement in both per-patient and per-slice analyses. The intraclass correllation of the T1 calculations by the automatic method vs reader 1 and reader 2 was respectively 0.86/0.56 and 0.74/0.49 in the per-patient/per-slice analyses, which were comparable to that between two expert readers (=0.72/0.58 in per-patient/per-slice analyses).ConclusionThe proposed FCN-based image processing platform allows fast and automatic analysis of myocardial native T1 mapping images mitigating the burden and observer-related variability of manual analysis.

Project description:The quantification of new or enlarged lesions from follow-up MRI scans is an important surrogate of clinical disease activity in patients with multiple sclerosis (MS). Not only is manual segmentation time consuming, but inter-rater variability is high. Currently, only a few fully automated methods are available. We address this gap in the field by employing a 3D convolutional neural network (CNN) with encoder-decoder architecture for fully automatic longitudinal lesion segmentation. Input data consist of two fluid attenuated inversion recovery (FLAIR) images (baseline and follow-up) per patient. Each image is entered into the encoder and the feature maps are concatenated and then fed into the decoder. The output is a 3D mask indicating new or enlarged lesions (compared to the baseline scan). The proposed method was trained on 1809 single point and 1444 longitudinal patient data sets and then validated on 185 independent longitudinal data sets from two different scanners. From the two validation data sets, manual segmentations were available from three experienced raters, respectively. The performance of the proposed method was compared to the open source Lesion Segmentation Toolbox (LST), which is a current state-of-art longitudinal lesion segmentation method. The mean lesion-wise inter-rater sensitivity was 62%, while the mean inter-rater number of false positive (FP) findings was 0.41 lesions per case. The two validated algorithms showed a mean sensitivity of 60% (CNN), 46% (LST) and a mean FP of 0.48 (CNN), 1.86 (LST) per case. Sensitivity and number of FP were not significantly different (p < 0.05) between the CNN and manual raters. New or enlarged lesions counted by the CNN algorithm appeared to be comparable with manual expert ratings. The proposed algorithm seems to outperform currently available approaches, particularly LST. The high inter-rater variability in case of manual segmentation indicates the complexity of identifying new or enlarged lesions. An automated CNN-based approach can quickly provide an independent and deterministic assessment of new or enlarged lesions from baseline to follow-up scans with acceptable reliability.

Project description:BackgroundCine Displacement Encoding with Stimulated Echoes (DENSE) facilitates the quantification of myocardial deformation, by encoding tissue displacements in the cardiovascular magnetic resonance (CMR) image phase, from which myocardial strain can be estimated with high accuracy and reproducibility. Current methods for analyzing DENSE images still heavily rely on user input, making this process time-consuming and subject to inter-observer variability. The present study sought to develop a spatio-temporal deep learning model for segmentation of the left-ventricular (LV) myocardium, as spatial networks often fail due to contrast-related properties of DENSE images.Methods2D + time nnU-Net-based models have been trained to segment the LV myocardium from DENSE magnitude data in short- and long-axis images. A dataset of 360 short-axis and 124 long-axis slices was used to train the networks, from a combination of healthy subjects and patients with various conditions (hypertrophic and dilated cardiomyopathy, myocardial infarction, myocarditis). Segmentation performance was evaluated using ground-truth manual labels, and a strain analysis using conventional methods was performed to assess strain agreement with manual segmentation. Additional validation was performed using an externally acquired dataset to compare the inter- and intra-scanner reproducibility with respect to conventional methods.ResultsSpatio-temporal models gave consistent segmentation performance throughout the cine sequence, while 2D architectures often failed to segment end-diastolic frames due to the limited blood-to-myocardium contrast. Our models achieved a DICE score of 0.83 ± 0.05 and a Hausdorff distance of 4.0 ± 1.1 mm for short-axis segmentation, and 0.82 ± 0.03 and 7.9 ± 3.9 mm respectively for long-axis segmentations. Strain measurements obtained from automatically estimated myocardial contours showed good to excellent agreement with manual pipelines, and remained within the limits of inter-user variability estimated in previous studies.ConclusionSpatio-temporal deep learning shows increased robustness for the segmentation of cine DENSE images. It provides excellent agreement with manual segmentation for strain extraction. Deep learning will facilitate the analysis of DENSE data, bringing it one step closer to clinical routine.

Project description:Importance:Retinopathy of prematurity (ROP) is a leading cause of childhood blindness worldwide. The decision to treat is primarily based on the presence of plus disease, defined as dilation and tortuosity of retinal vessels. However, clinical diagnosis of plus disease is highly subjective and variable. Objective:To implement and validate an algorithm based on deep learning to automatically diagnose plus disease from retinal photographs. Design, Setting, and Participants:A deep convolutional neural network was trained using a data set of 5511 retinal photographs. Each image was previously assigned a reference standard diagnosis (RSD) based on consensus of image grading by 3 experts and clinical diagnosis by 1 expert (ie, normal, pre-plus disease, or plus disease). The algorithm was evaluated by 5-fold cross-validation and tested on an independent set of 100 images. Images were collected from 8 academic institutions participating in the Imaging and Informatics in ROP (i-ROP) cohort study. The deep learning algorithm was tested against 8 ROP experts, each of whom had more than 10 years of clinical experience and more than 5 peer-reviewed publications about ROP. Data were collected from July 2011 to December 2016. Data were analyzed from December 2016 to September 2017. Exposures:A deep learning algorithm trained on retinal photographs. Main Outcomes and Measures:Receiver operating characteristic analysis was performed to evaluate performance of the algorithm against the RSD. Quadratic-weighted ? coefficients were calculated for ternary classification (ie, normal, pre-plus disease, and plus disease) to measure agreement with the RSD and 8 independent experts. Results:Of the 5511 included retinal photographs, 4535 (82.3%) were graded as normal, 805 (14.6%) as pre-plus disease, and 172 (3.1%) as plus disease, based on the RSD. Mean (SD) area under the receiver operating characteristic curve statistics were 0.94 (0.01) for the diagnosis of normal (vs pre-plus disease or plus disease) and 0.98 (0.01) for the diagnosis of plus disease (vs normal or pre-plus disease). For diagnosis of plus disease in an independent test set of 100 retinal images, the algorithm achieved a sensitivity of 93% with 94% specificity. For detection of pre-plus disease or worse, the sensitivity and specificity were 100% and 94%, respectively. On the same test set, the algorithm achieved a quadratic-weighted ? coefficient of 0.92 compared with the RSD, outperforming 6 of 8 ROP experts. Conclusions and Relevance:This fully automated algorithm diagnosed plus disease in ROP with comparable or better accuracy than human experts. This has potential applications in disease detection, monitoring, and prognosis in infants at risk of ROP.

Project description:Background Cardiac amyloidosis (CA) is a disease of interstitial myocardial infiltration, usually by light chains or transthyretin. We used diffusion tensor cardiovascular magnetic resonance (DT-CMR) to noninvasively assess the effects of amyloid infiltration on the cardiac microstructure. Methods DT-CMR was performed at diastole and systole in 20 CA, 11 hypertrophic cardiomyopathy, and 10 control subjects with calculation of mean diffusivity, fractional anisotropy, and sheetlet orientation (secondary eigenvector angle). Results Mean diffusivity was elevated and fractional anisotropy reduced in CA compared with both controls and hypertrophic cardiomyopathy (P<0.001). In CA, mean diffusivity was correlated with extracellular volume (r=0.68, P=0.004), and fractional anisotropy was inversely correlated with circumferential strain (r=-0.65, P=0.02). In CA, diastolic secondary eigenvector angle was elevated, and secondary eigenvector angle mobility was reduced compared with controls (both P<0.001). Diastolic secondary eigenvector angle was correlated with amyloid burden measured by extracellular volume in transthyretin, but not light chain amyloidosis. Conclusions DT-CMR can characterize the microstructural effects of amyloid infiltration and is a contrast-free method to identify the location and extent of the expanded disorganized myocardium. The diffusion biomarkers mean diffusivity and fractional anisotropy effectively discriminate CA from hypertrophic cardiomyopathy. DT-CMR demonstrated that failure of sheetlet relaxation in diastole correlated with extracellular volume in transthyretin, but not light chain amyloidosis. This indicates that different mechanisms may be responsible for impaired contractility in CA, with an amyloid burden effect in transthyretin, but an idiosyncratic effect in light chain amyloidosis. Consequently, DT-CMR offers a contrast-free tool to identify novel pathophysiology, improve diagnostics, and monitor disease through noninvasive microstructural assessment.

Project description:Brain tumors are a diverse group of neoplasms that are challenging to detect and classify due to their varying characteristics. Deep learning techniques have proven to be effective in tumor classification. However, there is a lack of studies that compare these techniques using a common methodology. This work aims to analyze the performance of convolutional neural networks in the classification of brain tumors. We propose a network consisting of a few convolutional layers, batch normalization, and max-pooling. Then, we explore recent deep architectures, such as VGG, ResNet, EfficientNet, or ConvNeXt. The study relies on two magnetic resonance imaging datasets with over 3000 images of three types of tumors -gliomas, meningiomas, and pituitary tumors-, as well as images without tumors. We determine the optimal hyperparameters of the networks using the training and validation sets. The training and test sets are used to assess the performance of the models from different perspectives, including training from scratch, data augmentation, transfer learning, and fine-tuning. The experiments are performed using the TensorFlow and Keras libraries in Python. We compare the accuracy of the models and analyze their complexity based on the capacity of the networks, their training times, and image throughput. Several networks achieve high accuracy rates on both datasets, with the best model achieving 98.7% accuracy, which is on par with state-of-the-art methods. The average precision for each type of tumor is 94.3% for gliomas, 93.8% for meningiomas, 97.9% for pituitary tumors, and 95.3% for images without tumors. VGG is the largest model with over 171 million parameters, whereas MobileNet and EfficientNetB0 are the smallest ones with 3.2 and 5.9 million parameters, respectively. These two neural networks are also the fastest to train with 23.7 and 25.4 seconds per epoch, respectively. On the other hand, ConvNext is the slowest model with 58.2 seconds per epoch. Our custom model obtained the highest image throughput with 234.37 images per second, followed by MobileNet with 226 images per second. ConvNext obtained the smallest throughput with 97.35 images per second. ResNet, MobileNet, and EfficientNet are the most accurate networks, with MobileNet and EfficientNet demonstrating superior performance in terms of complexity. Most models achieve the best accuracy using transfer learning followed by a fine-tuning step. However, data augmentation does not contribute to increasing the accuracy of the models in general.