Automatic Spatial Estimation of White Matter Hyperintensities Evolution in Brain MRI using Disease Evolution Predictor Deep Neural Networks

Abstract

<jats:title>Abstract</jats:title><jats:p>Previous studies have indicated that white matter hyperintensities (WMH), the main radiological feature of small vessel disease, may evolve (i.e., shrink, grow) or stay stable over a period of time. Predicting these changes are challenging because it involves some unknown clinical risk factors that leads to a non-deterministic prediction task. In this study, we propose a deep learning model to predict the evolution of WMH from baseline to follow-up (i.e., 1-year later), namely “Disease Evolution Predictor” (DEP) model, which can be adjusted to become a non-deterministic model. The DEP model receives a baseline image as input and produces a map called “Disease Evolution Map” (DEM), which represents the evolution of WMH from baseline to follow-up. Two DEP models are proposed, namely DEP-UResNet and DEP-GAN, which are representatives of the supervised (i.e., need expert-generated manual labels to generate the output) and unsupervised (i.e., do not require manual labels produced by experts) deep learning algorithms respectively. To simulate the non-deterministic and unknown parameters involved in WMH evolution, we modulate a Gaussian noise array to the DEP model as auxiliary input. This forces the DEP model to imitate a wider spectrum of alternatives in the prediction results. The alternatives of using other types of auxiliary input instead, such as baseline WMH and stroke lesion loads are also proposed and tested. Based on our experiments, the fully supervised machine learning scheme DEP-UResNet regularly performed better than the DEP-GAN which works in principle without using any expert-generated label (i.e., unsupervised). However, a semi-supervised DEP-GAN model, which uses probability maps produced by a supervised segmentation method in the learning process, yielded similar performances to the DEP-UResNet and performed best in the clinical evaluation. Furthermore, an ablation study showed that an auxiliary input, especially the Gaussian noise, improved the performance of DEP models compared to DEP models that lacked the auxiliary input regardless of the model’s architecture. To the best of our knowledge, this is the first extensive study on modelling WMH evolution using deep learning algorithms, which deals with the non-deterministic nature of WMH evolution.</jats:p>

Learning to synthesise the ageing brain without longitudinal data

“…Deep generative methods have also been used for this task. While Rachmadi et al (2019Rachmadi et al ( , 2020 and Wegmayr et al (2019) used formulations of Generative Adversarial Networks (GAN) (Goodfellow et al, 2014) to simulate brain changes, others (Ravi et al, 2019a) used a conditional adversarial autoencoder as the generative model, following a recent face ageing approach (Zhang et al, 2017). Irrespective of the model, these methods need longitudinal data, which limits their applicability.…”

“…In summary, most previous methods either built average atlases (Zhang et al, 2016;Huizinga et al, 2018;Ziegler et al, 2012;Serag et al, 2012), or required longitudinal data (Rachmadi et al, 2019(Rachmadi et al, , 2020Ravi et al, 2019a;Wegmayr et al, 2019) to simulate brain ageing. Other methods either did not consider subject identity (Bowles et al, 2018;Milana, 2017), or did not evaluate in detail morphological changes (Pawlowski et al, 2020;Zhao et al, 2019).…”

“…However, such learned phenomenological predictive models need to capture anatomical and physiological changes due to ageing and separate the factors that influence future state. Recently, deep generative models have been used to simulate and predict future degeneration of a human brain using existing scans (Ravi et al, 2019a;Rachmadi et al, 2019Rachmadi et al, , 2020. However, current methods require considerable amount of longitudinal data to sufficiently approximate an auto-regressive model.…”

Emerging role of white matter lesions in cerebrovascular disease

“…In comparison to qualitative visual rating scales, volumetry provides greater accuracy and precision (Gouw et al, 2008; Straaten et al, 2006). The use of machine learning and deep learning algorithms shows promise and can further improve the sensitivity and prognostic accuracy of volumetric analysis (Rachmadi et al, 2020). However, the advanced imaging and software required for volumetric analysis may not be easily accessible, potentially making its use impractical in routine clinical practice.…”

An Uncertainty-aware Hierarchical Probabilistic Network for Early Prediction, Quantification and Segmentation of Pulmonary Tumour Growth

“…Although high performances were reported (88% Dice score), the database was composed by only 18 subjects, in which all tumours always grew. In Rachmadi et al (2020), they compared different GAN networks to predict the evolution of white matter hyperintensities. They also demonstrated the potential of using GANs in a semi-supervised scheme, improving results of a deterministic U-ResNet (Zhang et al, 2018).…”