Microaneurysms segmentation with a U-Net based on recurrent residual convolutional neural network

MCA-UNet: multi-scale cross co-attentional U-Net for automatic medical image segmentation

Medical image segmentation is a challenging task due to the high variation in shape, size and position of infections or lesions in medical images. It is necessary to construct multi-scale representations to capture image contents from different scales. However, it is still challenging for U-Net with a simple skip connection to model the global multi-scale context. To overcome it, we proposed a dense skip-connection with cross co-attention in U-Net to solve the semantic gaps for an accurate automatic medical image segmentation. We name our method MCA-UNet, which enjoys two benefits: (1) it has a strong ability to model the multi-scale features, and (2) it jointly explores the spatial and channel attentions. The experimental results on the COVID-19 and IDRiD datasets suggest that our MCA-UNet produces more precise segmentation performance for the consolidation, ground-glass opacity (GGO), microaneurysms (MA) and hard exudates (EX). The source code of this work will be released via https://github.com/McGregorWwww/MCA-UNet/.

Association of Biomarker-Based Artificial Intelligence With Risk of Racial Bias in Retinal Images

Importance

Although race is a social construct, it is associated with variations in skin and retinal pigmentation. Image-based medical artificial intelligence (AI) algorithms that use images of these organs have the potential to learn features associated with self-reported race (SRR), which increases the risk of racially biased performance in diagnostic tasks; understanding whether this information can be removed, without affecting the performance of AI algorithms, is critical in reducing the risk of racial bias in medical AI.

Objective

To evaluate whether converting color fundus photographs to retinal vessel maps (RVMs) of infants screened for retinopathy of prematurity (ROP) removes the risk for racial bias.

Design, Setting, and Participants

The retinal fundus images (RFIs) of neonates with parent-reported Black or White race were collected for this study. A u-net, a convolutional neural network (CNN) that provides precise segmentation for biomedical images, was used to segment the major arteries and veins in RFIs into grayscale RVMs, which were subsequently thresholded, binarized, and/or skeletonized. CNNs were trained with patients' SRR labels on color RFIs, raw RVMs, and thresholded, binarized, or skeletonized RVMs. Study data were analyzed from July 1 to September 28, 2021.

Main Outcomes and Measures

Area under the precision-recall curve (AUC-PR) and area under the receiver operating characteristic curve (AUROC) at both the image and eye level for classification of SRR.

Results

A total of 4095 RFIs were collected from 245 neonates with parent-reported Black (94 [38.4%]; mean [SD] age, 27.2 [2.3] weeks; 55 majority sex [58.5%]) or White (151 [61.6%]; mean [SD] age, 27.6 [2.3] weeks, 80 majority sex [53.0%]) race. CNNs inferred SRR from RFIs nearly perfectly (image-level AUC-PR, 0.999; 95% CI, 0.999-1.000; infant-level AUC-PR, 1.000; 95% CI, 0.999-1.000). Raw RVMs were nearly as informative as color RFIs (image-level AUC-PR, 0.938; 95% CI, 0.926-0.950; infant-level AUC-PR, 0.995; 95% CI, 0.992-0.998). Ultimately, CNNs were able to learn whether RFIs or RVMs were from Black or White infants regardless of whether images contained color, vessel segmentation brightness differences were nullified, or vessel segmentation widths were uniform.

Conclusions and Relevance

Results of this diagnostic study suggest that it can be very challenging to remove information relevant to SRR from fundus photographs. As a result, AI algorithms trained on fundus photographs have the potential for biased performance in practice, even if based on biomarkers rather than raw images. Regardless of the methodology used for training AI, evaluating performance in relevant subpopulations is critical.

MESTrans: Multi-scale embedding spatial transformer for medical image segmentation

BACKGROUND AND OBJECTIVE

Transformers profiting from global information modeling derived from the self-attention mechanism have recently achieved remarkable performance in computer vision. In this study, a novel transformer-based medical image segmentation network called the multi-scale embedding spatial transformer (MESTrans) was proposed for medical image segmentation.

METHODS

First, a dataset called COVID-DS36 was created from 4369 computed tomography (CT) images of 36 patients from a partner hospital, of which 18 had COVID-19 and 18 did not. Subsequently, a novel medical image segmentation network was proposed, which introduced a self-attention mechanism to improve the inherent limitation of convolutional neural networks (CNNs) and was capable of adaptively extracting discriminative information in both global and local content. Specifically, based on U-Net, a multi-scale embedding block (MEB) and multi-layer spatial attention transformer (SATrans) structure were designed, which can dynamically adjust the receptive field in accordance with the input content. The spatial relationship between multi-level and multi-scale image patches was modeled, and the global context information was captured effectively. To make the network concentrate on the salient feature region, a feature fusion module (FFM) was established, which performed global learning and soft selection between shallow and deep features, adaptively combining the encoder and decoder features. Four datasets comprising CT images, magnetic resonance (MR) images, and H&E-stained slide images were used to assess the performance of the proposed network.

RESULTS

Experiments were performed using four different types of medical image datasets. For the COVID-DS36 dataset, our method achieved a Dice similarity coefficient (DSC) of 81.23%. For the GlaS dataset, 89.95% DSC and 82.39% intersection over union (IoU) were obtained. On the Synapse dataset, the average DSC was 77.48% and the average Hausdorff distance (HD) was 31.69 mm. For the I2CVB dataset, 92.3% DSC and 85.8% IoU were obtained.

CONCLUSIONS

The experimental results demonstrate that the proposed model has an excellent generalization ability and outperforms other state-of-the-art methods. It is expected to be a potent tool to assist clinicians in auxiliary diagnosis and to promote the development of medical intelligence technology.

Retinal Disease Detection Using Deep Learning Techniques: A Comprehensive Review

Millions of people are affected by retinal abnormalities worldwide. Early detection and treatment of these abnormalities could arrest further progression, saving multitudes from avoidable blindness. Manual disease detection is time-consuming, tedious and lacks repeatability. There have been efforts to automate ocular disease detection, riding on the successes of the application of Deep Convolutional Neural Networks (DCNNs) and vision transformers (ViTs) for Computer-Aided Diagnosis (CAD). These models have performed well, however, there remain challenges owing to the complex nature of retinal lesions. This work reviews the most common retinal pathologies, provides an overview of prevalent imaging modalities and presents a critical evaluation of current deep-learning research for the detection and grading of glaucoma, diabetic retinopathy, Age-Related Macular Degeneration and multiple retinal diseases. The work concluded that CAD, through deep learning, will increasingly be vital as an assistive technology. As future work, there is a need to explore the potential impact of using ensemble CNN architectures in multiclass, multilabel tasks. Efforts should also be expended on the improvement of model explainability to win the trust of clinicians and patients.

Improved detection accuracy of red lesions in retinal fundus images with superlearning approach

BACKGROUND

Diabetic Retinopathy (DR) is a serious consequence of diabetes that can result to permanent vision loss for a person. Diabetes-related vision impairment can be significantly avoided with timely screening and treatment in its initial phase. The earliest and the most noticeable indications on the surface of the retina are micro-aneurysm and haemorrhage, which appear as dark patches. Therefore, the automatic detection of retinopathy begins with the identification of all these dark lesions.

METHOD

In our study, we have developed a clinical knowledge based segmentation built on Early Treatment DR Study (ETDRS). ETDRS is a gold standard for identifying all red lesions using adaptive-thresholding approach followed by different pre-processing steps. The lesions are classified using super-learning approach to improve multi-class detection accuracy. Ensemble based super-learning approach finds optimal weights of base learners by minimizing the cross validated risk-function and it pledges the improved performance compared to base-learners predictions. For multi-class classification, a well informative feature-set based on colour, intensity, shape, size and texture, is developed. In this work, we have handled the data imbalance problem and compared the final accuracy with different synthetic data creation ratios.

RESULT

The suggested approach uses publicly available resources to perform quantitative assessments at lesions-level. The overall accuracy of red lesion segregation is 93.5%, which has increased to 97.88% when data imbalance problem is taken care-off.

CONCLUSION

The results of our system have achieved competitive performance compared with other modern approaches and handling of data imbalance further increases the performance of it.

Deep Learning with a Dataset Created Using Kanno Saitama Macro, a Self-Made Automatic Foveal Avascular Zone Extraction Program

The extraction of the foveal avascular zone (FAZ) from optical coherence tomography angiography (OCTA) images has been used in many studies in recent years due to its association with various ophthalmic diseases. In this study, we investigated the utility of a dataset for deep learning created using Kanno Saitama Macro (KSM), a program that automatically extracts the FAZ using swept-source OCTA. The test data included 40 eyes of 20 healthy volunteers. For training and validation, we used 257 eyes from 257 patients. The FAZ of the retinal surface image was extracted using KSM, and a dataset for FAZ extraction was created. Based on that dataset, we conducted a training test using a typical U-Net. Two examiners manually extracted the FAZ of the test data, and the results were used as gold standards to compare the Jaccard coefficients between examiners, and between each examiner and the U-Net. The Jaccard coefficient was 0.931 between examiner 1 and examiner 2, 0.951 between examiner 1 and the U-Net, and 0.933 between examiner 2 and the U-Net. The Jaccard coefficients were significantly better between examiner 1 and the U-Net than between examiner 1 and examiner 2 (p < 0.001). These data indicated that the dataset generated by KSM was as good as, if not better than, the agreement between examiners using the manual method. KSM may contribute to reducing the burden of annotation in deep learning.

Combining transfer learning with retinal lesion features for accurate detection of diabetic retinopathy

Diabetic retinopathy (DR) is a late microvascular complication of Diabetes Mellitus (DM) that could lead to permanent blindness in patients, without early detection. Although adequate management of DM via regular eye examination can preserve vision in in 98% of the DR cases, DR screening and diagnoses based on clinical lesion features devised by expert clinicians; are costly, time-consuming and not sufficiently accurate. This raises the requirements for Artificial Intelligent (AI) systems which can accurately detect DR automatically and thus preventing DR before affecting vision. Hence, such systems can help clinician experts in certain cases and aid ophthalmologists in rapid diagnoses. To address such requirements, several approaches have been proposed in the literature that use Machine Learning (ML) and Deep Learning (DL) techniques to develop such systems. However, these approaches ignore the highly valuable clinical lesion features that could contribute significantly to the accurate detection of DR. Therefore, in this study we introduce a framework called DR-detector that employs the Extreme Gradient Boosting (XGBoost) ML model trained via the combination of the features extracted by the pretrained convolutional neural networks commonly known as transfer learning (TL) models and the clinical retinal lesion features for accurate detection of DR. The retinal lesion features are extracted via image segmentation technique using the UNET DL model and captures exudates (EXs), microaneurysms (MAs), and hemorrhages (HEMs) that are relevant lesions for DR detection. The feature combination approach implemented in DR-detector has been applied to two common TL models in the literature namely VGG-16 and ResNet-50. We trained the DR-detector model using a training dataset comprising of 1,840 color fundus images collected from e-ophtha, retinal lesions and APTOS 2019 Kaggle datasets of which 920 images are healthy. To validate the DR-detector model, we test the model on external dataset that consists of 81 healthy images collected from High-Resolution Fundus (HRF) dataset and MESSIDOR-2 datasets and 81 images with DR signs collected from Indian Diabetic Retinopathy Image Dataset (IDRID) dataset annotated for DR by expert. The experimental results show that the DR-detector model achieves a testing accuracy of 100% in detecting DR after training it with the combination of ResNet-50 and lesion features and 99.38% accuracy after training it with the combination of VGG-16 and lesion features. More importantly, the results also show a higher contribution of specific lesion features toward the performance of the DR-detector model. For instance, using only the hemorrhages feature to train the model, our model achieves an accuracy of 99.38 in detecting DR, which is higher than the accuracy when training the model with the combination of all lesion features (89%) and equal to the accuracy when training the model with the combination of all lesions and VGG-16 features together. This highlights the possibility of using only the clinical features, such as lesions that are clinically interpretable, to build the next generation of robust artificial intelligence (AI) systems with great clinical interpretability for DR detection. The code of the DR-detector framework is available on GitHub at https://github.com/Janga-Lab/DR-detector and can be readily employed for detecting DR from retinal image datasets.

Multi-scale discriminative network for prostate cancer lesion segmentation in multiparametric MR images

PURPOSE

The accurate and reliable segmentation of prostate cancer (PCa) lesions using multiparametric magnetic resonance imaging (mpMRI) sequences, is crucial to the image-guided intervention and treatment of prostate disease. For PCa lesion segmentation, it is essential to reliably combine local and global information to retain the features of small targets at multiple scales. Therefore, this study proposes a multi-scale segmentation network with a cascading pyramid convolution module (CPCM) and a double-input channel attention module (DCAM) for the automated and accurate segmentation of PCa lesions using mpMRI.

METHODS

First, the region of interest was extracted from the data by clipping to enlarge the target region and reduce the background noise interference. Next, four CPCMs with large convolution kernels in their skip connection paths were designed to improve the feature extraction capability of the network for small targets. At the same time, a convolution decomposition was applied to reduce the computational complexity. Finally, the DCAM was adopted in the decoder to provide bottom-up semantic discriminative guidance; it can use the semantic information of the network's deep features to guide the shallow output of features with a higher discriminant ability. A residual refinement module (RRM) was also designed to strengthen the recognition ability of each stage. The feature maps of the skip connection and the decoder all go through the RRM.

RESULTS

For the Initiative for Collaborative Computer Vision Benchmarking (I2CVB) dataset, our proposed model achieved a Dice similarity coefficient (DSC) of 79.31% and an average boundary distance (ABD) of 4.15 mm. For the Prostate Multiparametric MRI (PROMM) dataset, our method greatly improved the DSC to 82.11% and obtained an ABD of 3.64 mm.

CONCLUSIONS

The experimental results of two different mpMRI prostate datasets demonstrate that our model is more accurate and reliable on small targets. In addition, it outperforms other state-of-the-art methods.

AOSLO-net: A Deep Learning-Based Method for Automatic Segmentation of Retinal Microaneurysms From Adaptive Optics Scanning Laser Ophthalmoscopy Images

Purpose

Accurate segmentation of microaneurysms (MAs) from adaptive optics scanning laser ophthalmoscopy (AOSLO) images is crucial for identifying MA morphologies and assessing the hemodynamics inside the MAs. Herein, we introduce AOSLO-net to perform automatic MA segmentation from AOSLO images of diabetic retinas.

Method

AOSLO-net is composed of a deep neural network based on UNet with a pretrained EfficientNet as the encoder. We have designed customized preprocessing and postprocessing policies for AOSLO images, including generation of multichannel images, de-noising, contrast enhancement, ensemble and union of model predictions, to optimize the MA segmentation. AOSLO-net is trained and tested using 87 MAs imaged from 28 eyes of 20 subjects with varying severity of diabetic retinopathy (DR), which is the largest available AOSLO dataset for MA detection. To avoid the overfitting in the model training process, we augment the training data by flipping, rotating, scaling the original image to increase the diversity of data available for model training.

Results

The validity of the model is demonstrated by the good agreement between the predictions of AOSLO-net and the MA masks generated by ophthalmologists and skillful trainees on 87 patient-specific MA images. Our results show that AOSLO-net outperforms the state-of-the-art segmentation model (nnUNet) both in accuracy (e.g., intersection over union and Dice scores), as well as computational cost.

Conclusions

We demonstrate that AOSLO-net provides high-quality of MA segmentation from AOSLO images that enables correct MA morphological classification.

Translational Relevance

As the first attempt to automatically segment retinal MAs from AOSLO images, AOSLO-net could facilitate the pathological study of DR and help ophthalmologists make disease prognoses.

RTNet: Relation Transformer Network for Diabetic Retinopathy Multi-Lesion Segmentation

Automatic diabetic retinopathy (DR) lesions segmentation makes great sense of assisting ophthalmologists in diagnosis. Although many researches have been conducted on this task, most prior works paid too much attention to the designs of networks instead of considering the pathological association for lesions. Through investigating the pathogenic causes of DR lesions in advance, we found that certain lesions are closed to specific vessels and present relative patterns to each other. Motivated by the observation, we propose a relation transformer block (RTB) to incorporate attention mechanisms at two main levels: a self-attention transformer exploits global dependencies among lesion features, while a cross-attention transformer allows interactions between lesion and vessel features by integrating valuable vascular information to alleviate ambiguity in lesion detection caused by complex fundus structures. In addition, to capture the small lesion patterns first, we propose a global transformer block (GTB) which preserves detailed information in deep network. By integrating the above blocks of dual-branches, our network segments the four kinds of lesions simultaneously. Comprehensive experiments on IDRiD and DDR datasets well demonstrate the superiority of our approach, which achieves competitive performance compared to state-of-the-arts.

A new detection model of microaneurysms based on improved FC-DenseNet

Diabetic retinopathy (DR) is a frequent vascular complication of diabetes mellitus and remains a leading cause of vision loss worldwide. Microaneurysm (MA) is usually the first symptom of DR that leads to blood leakage in the retina. Periodic detection of MAs will facilitate early detection of DR and reduction of vision injury. In this study, we proposed a novel model for the detection of MAs in fluorescein fundus angiography (FFA) images based on the improved FC-DenseNet, MAs-FC-DenseNet. FFA images were pre-processed by the Histogram Stretching and Gaussian Filtering algorithm to improve the quality of FFA images. Then, MA regions were detected by the improved FC-DenseNet. MAs-FC-DenseNet was compared against other FC-DenseNet models (FC-DenseNet56 and FC-DenseNet67) or the end-to-end models (DeeplabV3+ and PSPNet) to evaluate the detection performance of MAs. The result suggested that MAs-FC-DenseNet had higher values of evaluation metrics than other models, including pixel accuracy (PA), mean pixel accuracy (MPA), precision (Pre), recall (Re), F1-score (F1), and mean intersection over union (MIoU). Moreover, MA detection performance for MAs-FC-DenseNet was very close to the ground truth. Taken together, MAs-FC-DenseNet is a reliable model for rapid and accurate detection of MAs, which would be used for mass screening of DR patients.

Joint Learning of Multi-level Tasks for Diabetic Retinopathy Grading on Low-resolution Fundus Images

Diabetic retinopathy (DR) is a leading cause of permanent blindness among the working-age people. Automatic DR grading can help ophthalmologists make timely treatment for patients. However, the existing grading methods are usually trained with high resolution (HR) fundus images, such that the grading performance decreases a lot given low resolution (LR) images, which are common in clinic. In this paper, we mainly focus on DR grading with LR fundus images. According to our analysis on the DR task, we find that: 1) image super-resolution (ISR) can boost the performance of both DR grading and lesion segmentation; 2) the lesion segmentation regions of fundus images are highly consistent with pathological regions for DR grading. Based on our findings, we propose a convolutional neural network (CNN)-based method for joint learning of multi-level tasks for DR grading, called DeepMT-DR, which can simultaneously handle the low-level task of ISR, the mid-level task of lesion segmentation and the high-level task of disease severity classification on LR fundus images. Moreover, a novel task-aware loss is developed to encourage ISR to focus on the pathological regions for its subsequent tasks: lesion segmentation and DR grading. Extensive experimental results show that our DeepMT-DR method significantly outperforms other state-of-the-art methods for DR grading over three datasets. In addition, our method achieves comparable performance in two auxiliary tasks of ISR and lesion segmentation.

Coarse-to-fine multiplanar D-SEA UNet for automatic 3D carotid segmentation in CTA images

PURPOSE

Carotid artery atherosclerotic stenosis accounts for 18-25% of ischemic stroke. In the evaluation of carotid atherosclerotic lesions, the automatic, accurate and rapid segmentation of the carotid artery is a priority issue that needs to be addressed urgently. However, the carotid artery area occupies a small target in computed tomography angiography (CTA) images, which affect the segmentation accuracy.

METHODS

We proposed a coarse-to-fine segmentation pipeline with the Multiplanar D-SEA UNet to achieve fully automatic carotid artery segmentation on the entire 3D CTA images, and compared with other four neural networks (3D-UNet, RA-UNet, Isensee-UNet, Multiplanar-UNet) by assessing Dice, Jaccard similarity coefficient, sensitivity, area under the curve and average hausdorff distance.

RESULTS

Our proposed method can achieve a mean Dice score of 91.51% on the 68 neck CTA scans from Beijing Hospital, which remarkably outperforms state-of-the-art 3D image segmentation methods. And the C2F segmentation pipeline can effectively improve segmentation accuracy while avoiding resolution loss.

CONCLUSION

The proposed segmentation method can realize the fully automatic segmentation of the carotid artery and has robust performance with segmentation accuracy, which can be applied into plaque exfoliation and interventional surgery services. In addition, our method is easy to extend to other medical segmentation tasks with appropriate parameter settings.

Multiple Lesions Insertion: boosting diabetic retinopathy screening through Poisson editing

Deep neural networks have made incredible progress in many computer vision tasks, owing to access to a great amount of data. However, collecting ground truth for large medical image datasets is extremely inconvenient and difficult to implement in practical applications, due to high professional requirements. Synthesizing can generate meaningful supplement samples to enlarge the insufficient medical image dataset. In this study, we propose a new data augmentation method, Multiple Lesions Insertion (MLI), to simulate new diabetic retinopathy (DR) fundus images based on the healthy fundus images that insert real lesions, such as exudates, hemorrhages, microaneurysms templates, into new healthy fundus images with Poisson editing. The synthetic fundus images can be generated according to the clinical rules, i.e., in different DR grading fundus images, the number of exudates, hemorrhages, microaneurysms are different. The generated DR fundus images by our MLI method are realistic with the real texture features and rich details, without black spots, artifacts, and discontinuities. We first demonstrate the feasibility of this method in a DR computer-aided diagnosis (CAD) system, which judges whether the patient has transferred treatment or not. Our results indicate that the MLI method outperforms most of the traditional augmentation methods, i.e, oversampling, under-sampling, cropping, rotation, and adding other real sample methods in the DR screening task.

Three-dimensional image reconstruction based on improved U-net network for anatomy of pulmonary segmentectomy

Pulmonary segmentectomy is one of the advanced techniques in thoracic surgery, but it is difficult to understand and master because of its complex anatomical structure. The purpose of this study is to explore the application effect of three-dimensional (3D) image reconstruction based on an improved U-net network in the anatomy of thoracic surgery. In this study, a total of 40 standardization training residents of thoracic surgery in our hospital were randomly divided into two groups. The control group was taught by conventional thin-slice CT images, while the observation group was taught by 3D image reconstruction based on the improved U-net network. After the training process was completed, the teaching effect was compared between these two groups. Using the improved U-net network model, 3D reconstruction of pulmonary segments can be realized quickly. Compared with the control group, the individual and total objective scores in the observation group were higher. The satisfaction of learning interest, content understanding, clinical thinking mode, and understanding of operation process in the observation group was higher than that of the control group. From the results, we concluded that the 3D image reconstruction technology based on the improved U-net network could help students master the anatomical structure of pulmonary segments and improve their learning interest and clinical thinking ability.

SESV: Accurate Medical Image Segmentation by Predicting and Correcting Errors

Medical image segmentation is an essential task in computer-aided diagnosis. Despite their prevalence and success, deep convolutional neural networks (DCNNs) still need to be improved to produce accurate and robust enough segmentation results for clinical use. In this paper, we propose a novel and generic framework called Segmentation-Emendation-reSegmentation-Verification (SESV) to improve the accuracy of existing DCNNs in medical image segmentation, instead of designing a more accurate segmentation model. Our idea is to predict the segmentation errors produced by an existing model and then correct them. Since predicting segmentation errors is challenging, we design two ways to tolerate the mistakes in the error prediction. First, rather than using a predicted segmentation error map to correct the segmentation mask directly, we only treat the error map as the prior that indicates the locations where segmentation errors are prone to occur, and then concatenate the error map with the image and segmentation mask as the input of a re-segmentation network. Second, we introduce a verification network to determine whether to accept or reject the refined mask produced by the re-segmentation network on a region-by-region basis. The experimental results on the CRAG, ISIC, and IDRiD datasets suggest that using our SESV framework can improve the accuracy of DeepLabv3+ substantially and achieve advanced performance in the segmentation of gland cells, skin lesions, and retinal microaneurysms. Consistent conclusions can also be drawn when using PSPNet, U-Net, and FPN as the segmentation network, respectively. Therefore, our SESV framework is capable of improving the accuracy of different DCNNs on different medical image segmentation tasks.

Channel width optimized neural networks for liver and vessel segmentation in liver iron quantification

INTRODUCTION

MRI T2* relaxometry protocols are often used for Liver Iron Quantification in patients with hemochromatosis. Several methods exist to semi-automatically segment parenchyma and exclude vessels for this calculation.

PURPOSE

To determine if inclusion of multiple echoes inputs to Convolutional Neural Networks (CNN) improves automated liver and vessel segmentation in MRI T2* relaxometry protocols and to determine if the resultant segmentations agree with manual segmentations for liver iron quantification analysis.

METHODS

Multi echo Gradient Recalled Echo (GRE) MRI sequence for T2* relaxometry was performed for 79 exams on 31 patients with hemochromatosis for iron quantification analysis. 275 axial liver slices were manually segmented as ground truth masks. A batch normalized U-Net with variable input width to incorporate multiple echoes is used for segmentation, using DICE as the accuracy metric. ANOVA is used to evaluate significance of channel width changes in segmentation accuracy. Linear regression is used to model the relationship of channel width on segmentation accuracy. Liver segmentations are applied to relaxometry data to calculate liver T2* yielding liver iron concentration(LIC) derived from literature based calibration curves. Manual and CNN based LIC values are compared with Pearson correlation. Bland altman plots are used to visualize differences between manual and CNN based LIC values.

RESULTS

Performance metrics are tested on 55 hold out slices. Linear regression indicates that there is a monotonic increase of DICE with increasing channel depth (p = 0.001) with a slope of 3.61e-3. ANOVA indicates a significant increase segmentation accuracy over single channel starting at 3 channels. Incorporation of all channels results in an average DICE of 0.86, an average increase of 0.07 over single channel. The calculated LIC from CNN segmented livers agrees well with manual segmentation (R = 0.998, slope = 0.914, p«0.001), with an average absolute difference 0.27 ± 0.99 mg Fe/g or 1.34 ± 4.3%.

CONCLUSION

More input echoes yields higher model accuracy until the noise floor. Echos beyond the first three echo times in GRE based T2* relaxometry do not contribute significant information for segmentation of liver for LIC calculation. Deep learning models with three channel width allow for generalization of model to protocols of more than three echoes, effectively a universal requirement for relaxometry. Deep learning segmentations achieve a good accuracy compared with manual segmentations with minimal preprocessing. Liver iron values calculated from hand segmented liver and Neural network segmented liver were not statistically different from each other.