Deep Learning Segmentation of the Right Ventricle in Cardiac MRI: The M&Ms Challenge.

In recent years, several deep learning models have been proposed to accurately quantify and diagnose cardiac pathologies. These automated tools heavily rely on the accurate segmentation of cardiac structures in MRI images. However, segmentation of the right ventricle is challenging due to its highly complex shape and ill-defined borders. Hence, there is a need for new methods to handle such structure's geometrical and textural complexities, notably in the presence of pathologies such as Dilated Right Ventricle, Tricuspid Regurgitation, Arrhythmogenesis, Tetralogy of Fallot, and Inter-atrial Communication. The last MICCAI challenge on right ventricle segmentation was held in 2012 and included only 48 cases from a single clinical center. As part of the 12th Workshop on Statistical Atlases and Computational Models of the Heart (STACOM 2021), the M&Ms-2 challenge was organized to promote the interest of the research community around right ventricle segmentation in multi-disease, multi-view, and multi-center cardiac MRI. Three hundred sixty CMR cases, including short-axis and long-axis 4-chamber views, were collected from three Spanish hospitals using nine different scanners from three different vendors, and included a diverse set of right and left ventricle pathologies. The solutions provided by the participants show that nnU-Net achieved the best results overall. However, multi-view approaches were able to capture additional information, highlighting the need to integrate multiple cardiac diseases, views, scanners, and acquisition protocols to produce reliable automatic cardiac segmentation algorithms.

MyoPS: A benchmark of myocardial pathology segmentation combining three-sequence cardiac magnetic resonance images.

Assessment of myocardial viability is essential in diagnosis and treatment management of patients suffering from myocardial infarction, and classification of pathology on the myocardium is the key to this assessment. This work defines a new task of medical image analysis, i.e., to perform myocardial pathology segmentation (MyoPS) combining three-sequence cardiac magnetic resonance (CMR) images, which was first proposed in the MyoPS challenge, in conjunction with MICCAI 2020. Note that MyoPS refers to both myocardial pathology segmentation and the challenge in this paper. The challenge provided 45 paired and pre-aligned CMR images, allowing algorithms to combine the complementary information from the three CMR sequences for pathology segmentation. In this article, we provide details of the challenge, survey the works from fifteen participants and interpret their methods according to five aspects, i.e., preprocessing, data augmentation, learning strategy, model architecture and post-processing. In addition, we analyze the results with respect to different factors, in order to examine the key obstacles and explore the potential of solutions, as well as to provide a benchmark for future research. The average Dice scores of submitted algorithms were 0.614±0.231 and 0.644±0.153 for myocardial scars and edema, respectively. We conclude that while promising results have been reported, the research is still in the early stage, and more in-depth exploration is needed before a successful application to the clinics. MyoPS data and evaluation tool continue to be publicly available upon registration via its homepage (www.sdspeople.fudan.edu.cn/zhuangxiahai/0/myops20/).

Automatic uncertainty-based quality controlled T1 mapping and ECV analysis from native and post-contrast cardiac T1 mapping images using Bayesian vision transformer.

Deep learning-based methods for cardiac MR segmentation have achieved state-of-the-art results. However, these methods can generate incorrect segmentation results which can lead to wrong clinical decisions in the downstream tasks. Automatic and accurate analysis of downstream tasks, such as myocardial tissue characterization, is highly dependent on the quality of the segmentation results. Therefore, it is of paramount importance to use quality control methods to detect the failed segmentations before further analysis. In this work, we propose a fully automatic uncertainty-based quality control framework for T1 mapping and extracellular volume (ECV) analysis. The framework consists of three parts. The first one focuses on segmentation of cardiac structures from a native and post-contrast T1 mapping dataset (n=295) using a Bayesian Swin transformer-based U-Net. In the second part, we propose a novel uncertainty-based quality control (QC) to detect inaccurate segmentation results. The QC method utilizes image-level uncertainty features as input to a random forest-based classifier/regressor to determine the quality of the segmentation outputs. The experimental results from four different types of segmentation results show that the proposed QC method achieves a mean area under the ROC curve (AUC) of 0.927 on binary classification and a mean absolute error (MAE) of 0.021 on Dice score regression, significantly outperforming other state-of-the-art uncertainty based QC methods. The performance gap is notably higher in predicting the segmentation quality from poor-performing models which shows the robustness of our method in detecting failed segmentations. After the inaccurate segmentation results are detected and rejected by the QC method, in the third part, T1 mapping and ECV values are computed automatically to characterize the myocardial tissues of healthy and cardiac pathological cases. The native myocardial T1 and ECV values computed from automatic and manual segmentations show an excellent agreement yielding Pearson coefficients of 0.990 and 0.975 (on the combined validation and test sets), respectively. From the results, we observe that the automatically computed myocardial T1 and ECV values have the ability to characterize myocardial tissues of healthy and cardiac diseases like myocardial infarction, amyloidosis, Tako-Tsubo syndrome, dilated cardiomyopathy, and hypertrophic cardiomyopathy.

An Improved 3D Deep Learning-Based Segmentation of Left Ventricular Myocardial Diseases from Delayed-Enhancement MRI with Inclusion and Classification Prior Information U-Net (ICPIU-Net).

Accurate segmentation of the myocardial scar may supply relevant advancements in predicting and controlling deadly ventricular arrhythmias in subjects with cardiovascular disease. In this paper, we propose the architecture of inclusion and classification of prior information U-Net (ICPIU-Net) to efficiently segment the left ventricle (LV) myocardium, myocardial infarction (MI), and microvascular-obstructed (MVO) tissues from late gadolinium enhancement magnetic resonance (LGE-MR) images. Our approach was developed using two subnets cascaded to first segment the LV cavity and myocardium. Then, we used inclusion and classification constraint networks to improve the resulting segmentation of the diseased regions within the pre-segmented LV myocardium. This network incorporates the inclusion and classification information of the LGE-MRI to maintain topological constraints of pathological areas. In the testing stage, the outputs of each segmentation network obtained with specific estimated parameters from training were fused using the majority voting technique for the final label prediction of each voxel in the LGE-MR image. The proposed method was validated by comparing its results to manual drawings by experts from 50 LGE-MR images. Importantly, compared to various deep learning-based methods participating in the EMIDEC challenge, the results of our approach have a more significant agreement with manual contouring in segmenting myocardial diseases.

Automatic classification of tissues on pelvic MRI based on relaxation times and support vector machine.

Tissue segmentation and classification in MRI is a challenging task due to a lack of signal intensity standardization. MRI signal is dependent on the acquisition protocol, the coil profile, the scanner type, etc. While we can compute quantitative physical tissue properties independent of the hardware and the sequence parameters, it is still difficult to leverage these physical properties to segment and classify pelvic tissues. The proposed method integrates quantitative MRI values (T1 and T2 relaxation times and pure synthetic weighted images) and machine learning (Support Vector Machine (SVM)) to segment and classify tissues in the pelvic region, i.e.: fat, muscle, prostate, bone marrow, bladder, and air. Twenty-two men with a mean age of 30±14 years were included in this prospective study. The images were acquired with a 3 Tesla MRI scanner. An inversion recovery-prepared turbo spin echo sequence was used to obtain T1-weighted images at different inversion times with a TR of 14000 ms. A 32-echo spin echo sequence was used to obtain the T2-weighted images at different echo times with a TR of 5000 ms. T1 and T2 relaxation times, synthetic T1- and T2-weighted images and anatomical probabilistic maps were calculated and used as input features of a SVM for segmenting and classifying tissues within the pelvic region. The mean SVM classification accuracy across subjects was calculated for the different tissues: prostate (94.2%), fat (96.9%), muscle (95.8%), bone marrow (91%) and bladder (82.1%) indicating an excellent classification performance. However, the segmentation and classification for air (within the rectum) may not always be successful (mean SVM accuracy 47.5%) due to the lack of air data in the training and testing sets. Our findings suggest that SVM can reliably segment and classify tissues in the pelvic region.

Semi-automatic detection of myocardial trabeculation using cardiovascular magnetic resonance: correlation with histology and reproducibility in a mouse model of non-compaction.

BACKGROUND: The definition of left ventricular (LV) non-compaction is controversial, and discriminating between normal and excessive LV trabeculation remains challenging. Our goal was to quantify LV trabeculation on cardiovascular magnetic resonance (CMR) images in a genetic mouse model of non-compaction using a dedicated semi-automatic software package and to compare our results to the histology used as a gold standard. METHODS: Adult mice with ventricular non-compaction were generated by conditional trabecular deletion of Nkx2-5. Thirteen mice (5 controls, 8 Nkx2-5 mutants) were included in the study. Cine CMR series were acquired in the mid LV short axis plane (resolution 0.086 × 0.086x1mm3) (11.75 T). In a sub set of 6 mice, 5 to 7 cine CMR were acquired in LV short axis to cover the whole LV with a lower resolution (0.172 × 0.172x1mm3). We used semi-automatic software to quantify the compacted mass (Mc), the trabeculated mass (Mt) and the percentage of trabeculation (Mt/Mc) on all cine acquisitions. After CMR all hearts were sliced along the short axis and stained with eosin, and histological LV contouring was performed manually, blinded from the CMR results, and Mt, Mc and Mt/Mc were quantified. Intra and interobserver reproducibility was evaluated by computing the intra class correlation coefficient (ICC). RESULTS: Whole heart acquisition showed no statistical significant difference between trabeculation measured at the basal, midventricular and apical parts of the LV. On the mid-LV cine CMR slice, the median Mt was 0.92 mg (range 0.07-2.56 mg), Mc was 12.24 mg (9.58-17.51 mg), Mt/Mc was 6.74% (0.66-17.33%). There was a strong correlation between CMR and the histology for Mt, Mc and Mt/ Mc with respectively: r2 = 0.94 (p < 0.001), r2 = 0.91 (p < 0.001), r2 = 0.83 (p < 0.001). Intra- and interobserver reproducibility was 0.97 and 0.8 for Mt; 0.98 and 0.97 for Mc; 0.96 and 0.72 for Mt/Mc, respectively and significantly more trabeculation was observed in the Mc Mutant mice than the controls. CONCLUSION: The proposed semi-automatic quantification software is accurate in comparison to the histology and reproducible in evaluating Mc, Mt and Mt/ Mc on cine CMR.

Automatic deformable PET/MRI registration for preclinical studies based on B-splines and non-linear intensity transformation.

PET images deliver functional data, whereas MRI images provide anatomical information. Merging the complementary information from these two modalities is helpful in oncology. Alignment of PET/MRI images requires the use of multi-modal registration methods. Most of existing PET/MRI registration methods have been developed for humans and few works have been performed for small animal images. We proposed an automatic tool allowing PET/MRI registration for pre-clinical study based on a two-level hierarchical approach. First, we applied a non-linear intensity transformation to the PET volume to enhance. The global deformation is modeled by an affine transformation initialized by a principal component analysis. A free-form deformation based on B-splines is then used to describe local deformations. Normalized mutual information is used as voxel-based similarity measure. To validate our method, CT images acquired simultaneously with the PET on tumor-bearing mice were used. Results showed that the proposed algorithm outperformed affine and deformable registration techniques without PET intensity transformation with an average error of 0.72 ± 0.44 mm. The optimization time was reduced by 23% due to the introduction of robust initialization. In this paper, an automatic deformable PET-MRI registration algorithm for small animals is detailed and validated. Graphical abstract ᅟ.

What are normal relaxation times of tissues at 3 T?

The T1 and T2 relaxation times are the basic parameters behind magnetic resonance imaging. The accurate knowledge of the T1 and T2 values of tissues allows to perform quantitative imaging and to develop and optimize magnetic resonance sequences. A vast extent of methods and sequences has been developed to calculate the T1 and T2 relaxation times of different tissues in diverse centers. Surprisingly, a wide range of values has been reported for similar tissues (e.g. T1 of white matter from 699 to 1735ms and T2 of fat from 41 to 371ms), and the true values that represent each specific tissue are still unclear, which have deterred their common use in clinical diagnostic imaging. This article presents a comprehensive review of the reported relaxation times in the literature in vivo at 3T for a large span of tissues. It gives a detailed analysis of the different methods and sequences used to calculate the relaxation times, and it explains the reasons of the spread of reported relaxation times values in the literature.

A novel alternative to classify tissues from T 1 and T 2 relaxation times for prostate MRI.

OBJECTIVE: To segment and classify the different attenuation regions from MRI at the pelvis level using the T 1 and T 2 relaxation times and anatomical knowledge as a first step towards the creation of PET/MR attenuation maps. MATERIALS AND METHODS: Relaxation times were calculated by fitting the pixel-wise intensities of acquired T 1- and T 2-weighted images from eight men with inversion-recovery and multi-echo multi-slice spin-echo sequences. A decision binary tree based on relaxation times was implemented to segment and classify fat, muscle, prostate, and air (within the body). Connected component analysis and an anatomical knowledge-based procedure were implemented to localize the background and bone. RESULTS: Relaxation times at 3 T are reported for fat (T 1 = 385 ms, T 2 = 121 ms), muscle (T 1 = 1295 ms, T 2 = 40 ms), and prostate (T 1 = 1700 ms, T 2 = 80 ms). The relaxation times allowed the segmentation-classification of fat, prostate, muscle, and air, and combined with anatomical knowledge, they allowed classification of bone. The good segmentation-classification of prostate [mean Dice similarity score (mDSC) = 0.70] suggests a viable implementation in oncology and that of fat (mDSC = 0.99), muscle (mDSC = 0.99), and bone (mDSCs = 0.78) advocates for its implementation in PET/MR attenuation correction. CONCLUSION: Our method allows the segmentation and classification of the attenuation-relevant structures required for the generation of the attenuation map of PET/MR systems in prostate imaging: air, background, bone, fat, muscle, and prostate.

Semiautomatic detection of myocardial contours in order to investigate normal values of the left ventricular trabeculated mass using MRI.

PURPOSE: To propose, assess, and validate a semiautomatic method allowing rapid and reproducible measurement of trabeculated and compacted left ventricular (LV) masses from cardiac magnetic resonance imaging (MRI). MATERIALS AND METHODS: We developed a method to automatically detect noncompacted, endocardial, and epicardial contours. Papillary muscles were segmented using semiautomatic thresholding and were included in the compacted mass. Blood was removed from trabeculae using the same threshold tool. Trabeculated, compacted masses and ratio of noncompacted to compacted (NC:C) masses were computed. Preclinical validation was performed on four transgenic mice with hypertrabeculation of the LV (high-resolution cine imaging, 11.75T). Then analysis was performed on normal cine-MRI examinations (steady-state free precession [SSFP] sequences, 1.5T or 3T) obtained from 60 healthy participants (mean age 49 ± 16 years) with 10 men and 10 women for each of the following age groups: [20,39], [40,59], and [60,79]. Interobserver and interexamination segmentation reproducibility was assessed by using Bland-Altman analysis and by computing the correlation coefficient. RESULTS: In normal participants, noncompacted and compacted masses were 6.29 ± 2.03 g/m(2) and 62.17 ± 11.32 g/m(2) , respectively. The NC:C mass ratio was 10.26 ± 3.27%. Correlation between the two observers was from 0.85 for NC:C ratio to 0.99 for end-diastolic volume (P < 10(-5) ). The bias between the two observers was -1.06 ± 1.02 g/m(2) for trabeculated mass, -1.41 ± 2.78 g/m(2) for compacted mass, and -1.51 ± 1.77% for NC:C ratio. CONCLUSION: We propose a semiautomatic method based on region growing, active contours, and thresholding to calculate the NC:C mass ratio. This method is highly reproducible and might help in the diagnosis of LV noncompaction cardiomyopathy. J. Magn. Reson. Imaging 2016;43:1398-1406.

Assessing spatial resolution versus sensitivity from laser speckle contrast imaging: application to frequency analysis.

For blood perfusion monitoring, laser speckle contrast (LSC) imaging is a recent non-contact technique that has the characteristic of delivering noise-like speckled images. To exploit LSC images for quantitative physiological measurements, we developed an approach that implements controlled spatial averaging to reduce the detrimental impact of the noise and improve measurement sensitivity. By this approach, spatial resolution and measurement sensitivity can be traded-off in a flexible way depending on the quantitative prospect of the study. As an application, detectability of the cardiac activity from LSC images of forearm using power spectrum analysis is studied through the construction of spatial activity maps offering a window on the blood flow perfusion and its regional distribution. Comparisons with results obtained with signals of laser Doppler flowmetry probes are performed.

Laser speckle contrast imaging accurately measures blood flow over moving skin surfaces.

Cutaneous blood flow (CBF) can be assessed non-invasively with lasers. Unfortunately, movement artefacts in the laser skin signal (LS(sk)) might sometimes compromise the interpretation of the data. To date, no method is available to remove movement artefacts point-by-point. Using a laser speckle contrast imager, we simultaneously recorded LS(sk) and the signal backscattered from an adjacent opaque surface (LS(os)). The completion of a first protocol allowed a definition of a simple equation to calculate the CBF from movement artefact-affected traces of LS(sk) and LS(os). We then recorded LS(sk) and LS(os) before, during and for 5 min after the tourniquet ischemia, both when subjects (n=8) were immobile or submitted to external passive movements of random intensity throughout the test. The typical post-occlusive reactive hyperemia trace was not identifiable within the LS(sk) recordings, with LS(sk) being 2 to 3 times higher during movements than in the immobile situation. After the calculation of CBF, traces in the immobile versus movement conditions were comparable, with the "r" cross-correlation coefficient being 0.930+/-0.010. Our method might facilitate future investigations in microvascular physiology and pathophysiology, specifically in subjects who have frequent or continuous involuntary movements.