Generalisability of fetal ultrasound deep learning models to low-resource imaging settings in five African countries.

Most artificial intelligence (AI) research and innovations have concentrated in high-income countries, where imaging data, IT infrastructures and clinical expertise are plentiful. However, slower progress has been made in limited-resource environments where medical imaging is needed. For example, in Sub-Saharan Africa, the rate of perinatal mortality is very high due to limited access to antenatal screening. In these countries, AI models could be implemented to help clinicians acquire fetal ultrasound planes for the diagnosis of fetal abnormalities. So far, deep learning models have been proposed to identify standard fetal planes, but there is no evidence of their ability to generalise in centres with low resources, i.e. with limited access to high-end ultrasound equipment and ultrasound data. This work investigates for the first time different strategies to reduce the domain-shift effect arising from a fetal plane classification model trained on one clinical centre with high-resource settings and transferred to a new centre with low-resource settings. To that end, a classifier trained with 1792 patients from Spain is first evaluated on a new centre in Denmark in optimal conditions with 1008 patients and is later optimised to reach the same performance in five African centres (Egypt, Algeria, Uganda, Ghana and Malawi) with 25 patients each. The results show that a transfer learning approach for domain adaptation can be a solution to integrate small-size African samples with existing large-scale databases in developed countries. In particular, the model can be re-aligned and optimised to boost the performance on African populations by increasing the recall to [Formula: see text] and at the same time maintaining a high precision across centres. This framework shows promise for building new AI models generalisable across clinical centres with limited data acquired in challenging and heterogeneous conditions and calls for further research to develop new solutions for the usability of AI in countries with fewer resources and, consequently, in higher need of clinical support.

Light-Weight Localization and Scale-Independent Multi-gate UNET Segmentation of Left and Right Ventricles in MRI Images.

PURPOSE: Heart segmentation in cardiac magnetic resonance images is heavily used during the assessment of left ventricle global function. Automation of the segmentation is crucial to standardize the analysis. This study aims at developing a CNN-based framework to aid the clinical measurements of the left ventricle and right ventricle in cardiac magnetic resonance images. METHODS: We propose a fully automated framework for localization and segmentation of the left ventricle and right ventricle in both short- and long-axis views from cardiac magnetic resonance images. The localization module utilizes a light-weight model that detects the region of interest and feeds it to the segmentation model. Also, we propose the Multi-Gate block as an extension to the UNet to boost the segmentation performance by aggregating multi-scale features. Comparison between our proposed method and the baseline UNet was performed to show the gain in the overall performance. The reliability of the model was assessed by testing the method against cardiac magnetic resonance images with different levels of noise and deformations. RESULTS: Heart localization accuracy was 0.59 and 1.75 pixels in both short- and long-axis views respectively. Left and right ventricle blood-pool segmentation Dice was (0.93, 0.90) in end-systole and (0.97, 0.95) in end-diastole. The left ventricle myocardium was segmented accurately with Dice of 0.91 and 0.90 in end-systole and end-diastole respectively. Left ventricle ejection fraction was found to be highly correlated with the gold standard with r = 0.987. Moreover, the proposed pipeline is fast, achieving 0.002 sec per image on average. CONCLUSION: Adding the Multi-Gate Dilated Inception Block has boosted the performance of UNet architecture and has shown generalization ability when tested on noisy and deformed cardiac magnetic resonance images. The proposed method has proven its wide applicability and reliability for heart detection when tested on different datasets.

Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation: The M&Ms Challenge.

The emergence of deep learning has considerably advanced the state-of-the-art in cardiac magnetic resonance (CMR) segmentation. Many techniques have been proposed over the last few years, bringing the accuracy of automated segmentation close to human performance. However, these models have been all too often trained and validated using cardiac imaging samples from single clinical centres or homogeneous imaging protocols. This has prevented the development and validation of models that are generalizable across different clinical centres, imaging conditions or scanner vendors. To promote further research and scientific benchmarking in the field of generalizable deep learning for cardiac segmentation, this paper presents the results of the Multi-Centre, Multi-Vendor and Multi-Disease Cardiac Segmentation (M&Ms) Challenge, which was recently organized as part of the MICCAI 2020 Conference. A total of 14 teams submitted different solutions to the problem, combining various baseline models, data augmentation strategies, and domain adaptation techniques. The obtained results indicate the importance of intensity-driven data augmentation, as well as the need for further research to improve generalizability towards unseen scanner vendors or new imaging protocols. Furthermore, we present a new resource of 375 heterogeneous CMR datasets acquired by using four different scanner vendors in six hospitals and three different countries (Spain, Canada and Germany), which we provide as open-access for the community to enable future research in the field.

A computed tomography-based planning tool for predicting difficulty of minimally invasive aortic valve replacement.

OBJECTIVES: Minimally invasive aortic valve replacement has proven its value over the last decade by its significant advancement and reduction in mortality, morbidity and admission time. However, minimally invasive aortic valve replacement is associated with some on-site difficulties such as limited aortic annulus exposure. Currently, computed tomography scans are used to evaluate the anatomical relationship among the intercostal spaces, ascending aorta and aortic valve prior to surgery. We hypothesized that quantitative measurements of access distance and access angle are associated with outcome and access difficulty. METHODS: We introduce a novel minimally invasive aortic valve replacement planning prototype that allows automatic measurements of access angle, access distance and aortic annulus dimensions. The prototype visualizes these measurements on the chest cage as ISO contours. The association of these measures with outcome parameters such as extracorporeal circulation time, aortic cross-clamping time and access difficulty score was assessed. We included 14 patients who received a new valve by ministernotomy. RESULTS: The mean access angle was 40.3 ± 5.1°. It was strongly associated with aortic cross-clamping time (Pearson correlation coefficient = 0.60, P = 0.02) and access difficulty score (Spearman's rank correlation coefficient = 0.57, P = 0.03). Access angles were significantly different between easy and difficult access groups (P = 0.03). There was no significant association between access distance and outcome parameters. CONCLUSIONS: Access angle is strongly associated with procedure complexity. The automated presentation of this measure suggests added value of the prototype in clinical practice.

Dynamics of the aortic annulus in 4D CT angiography for transcatheter aortic valve implantation patients.

BACKGROUND: Transcatheter aortic valve implantation (TAVI) is a well-established treatment for patients with severe aortic valve stenosis. This procedure requires pre-operative planning by assessment of aortic dimensions on CT Angiography (CTA). It is well-known that the aortic root dimensions vary over the heart cycle. However, sizing is commonly performed at either mid-systole or end-diastole only, which has resulted in an inadequate understanding of its full dynamic behavior. STUDY GOAL: We studied the variation in annulus measurements during the cardiac cycle and determined if this variation is dependent on the amount of calcification at the annulus. METHODS: We measured and compared aortic root annular dimensions and calcium volume in CTA acquisitions at 10 cardiac cycle phases in 51 aortic stenosis patients. Sub-group analysis was performed based on the volume of calcium by splitting the population into mildly and severely calcified valves subgroups. RESULTS: For most annulus measurements, the largest differences were found between 10% and 70 to 80% cardiac cycle phases. Mean difference (±standard deviation) in annular minimum diameter, maximum diameter, area, and aspect ratio between mid-systole and end-diastole phases were 1.0 ± 0.29 mm (p = 0.065), 0.30 ± 0.24 mm (p = 0.7), 24.1 ± 7.6 mm2 (p < 0.001), and 0.041 ± 0.012 (p = 0.039) respectively. Calcium volume measurements varied strongly during the cardiac cycle. The dynamic annulus area was behaving differently between mildly and severely calcified subgroups (p = 0.02). Furthermore, patients with severe aortic calcification were associated with larger annulus diameters. CONCLUSION: There is a significant variation of annulus area and calcium volume measurement during the cardiac cycle. In our measurements, only the dynamic variation of the annulus area is dependent on the severity of the aortic calcification. For TAVI candidates, the annulus area is significantly larger in mid-systole compared to end-diastole.

Automated CTA based measurements for planning support of minimally invasive aortic valve replacement surgery.

Minimally invasive aortic valve replacement (mini-AVR) procedures are a valuable alternative to conventional open heart surgery. Currently, planning of mini-AVR consists of selection of the intercostal space closest to the sinotubular junction on preoperative computer tomography images. We developed an automated algorithm detecting the sinotubular junction (STJ) and intercostal spaces for finding the optimal incision location. The accuracy of the STJ detection was assessed by comparison with manual delineation by measuring the Euclidean distance between the manually and automatically detected points. In all 20 patients, the intercostal spaces were accurately detected. The median distance between automated and manually detected STJ locations was 1.4 [IQR= 0.91-4.7] mm compared to the interobserver variation of 1.0 [IQR= 0.54-1.3] mm. For 60% of patients, the fourth intercostal space was the closest to the STJ. The proposed algorithm is the first automated approach for detecting optimal incision location and has the potential to be implemented in clinical practice for planning of various mini-AVR procedures.

Automatic aortic root landmark detection in CTA images for preprocedural planning of transcatheter aortic valve implantation.

Transcatheter aortic valve implantation is currently a well-established minimal invasive treatment option for patients with severe aortic valve stenosis. CT Angiography is used for the pre-operative planning and sizing of the prosthesis. To reduce the inconsistency in sizing due to interobserver variability, we introduce and evaluate an automatic aortic root landmarks detection method to determine the sizing parameters. The proposed algorithm detects the sinotubular junction, two coronary ostia, and three valvular hinge points on a segmented aortic root surface. Using these aortic root landmarks, the automated method determines annulus radius, annulus orientation, and distance from annulus plane to right and left coronary ostia. Validation is performed by the comparison with manual measurements of two observers for 40 CTA image datasets. Detection of landmarks showed high accuracy where the mean distance between the automatically detected and reference landmarks was 2.81 ± 2.08 mm, comparable to the interobserver variation of 2.67 ± 2.52 mm. The mean annulus to coronary ostium distance was 16.9 ± 3.3 and 17.1 ± 3.3 mm for the automated and the reference manual measurements, respectively, with a mean paired difference of 1.89 ± 1.71 mm and interobserver mean paired difference of 1.38 ± 1.52 mm. Automated detection of aortic root landmarks enables automated sizing with good agreement with manual measurements, which suggests applicability of the presented method in current clinical practice.

A collaborative resource to build consensus for automated left ventricular segmentation of cardiac MR images.

A collaborative framework was initiated to establish a community resource of ground truth segmentations from cardiac MRI. Multi-site, multi-vendor cardiac MRI datasets comprising 95 patients (73 men, 22 women; mean age 62.73±11.24years) with coronary artery disease and prior myocardial infarction, were randomly selected from data made available by the Cardiac Atlas Project (Fonseca et al., 2011). Three semi- and two fully-automated raters segmented the left ventricular myocardium from short-axis cardiac MR images as part of a challenge introduced at the STACOM 2011 MICCAI workshop (Suinesiaputra et al., 2012). Consensus myocardium images were generated based on the Expectation-Maximization principle implemented by the STAPLE algorithm (Warfield et al., 2004). The mean sensitivity, specificity, positive predictive and negative predictive values ranged between 0.63 and 0.85, 0.60 and 0.98, 0.56 and 0.94, and 0.83 and 0.92, respectively, against the STAPLE consensus. Spatial and temporal agreement varied in different amounts for each rater. STAPLE produced high quality consensus images if the region of interest was limited to the area of discrepancy between raters. To maintain the quality of the consensus, an objective measure based on the candidate automated rater performance distribution is proposed. The consensus segmentation based on a combination of manual and automated raters were more consistent than any particular rater, even those with manual input. The consensus is expected to improve with the addition of new automated contributions. This resource is open for future contributions, and is available as a test bed for the evaluation of new segmentation algorithms, through the Cardiac Atlas Project (www.cardiacatlas.org).