A Deep Learning Approach to Predict Abdominal Aortic Aneurysm Expansion Using Longitudinal Data

Abstract

An abdominal aortic aneurysm (AAA) is a gradual enlargement of the aorta that can cause a life-threatening event when a rupture occurs. Aneurysmal geometry has been proved to be a critical factor in determining when to surgically treat AAAs, but, it is challenging to predict the patient-specific evolution of an AAA with biomechanical or statistical models. The recent success of deep learning in biomedical engineering shows promise for predictive medicine. However, a deep learning model requires a large dataset, which limits its application to the prediction of the patient-specific AAA expansion. In order to cope with the limited medical follow-up dataset of AAAs, a novel technique combining a physical computational model with a deep learning model is introduced to predict the evolution of AAAs. First, a vascular Growth and Remodeling (G&R) computational model, which is able to capture the variations of actual patient AAA geometries, is employed to generate a limited in silico dataset. Second, the Probabilistic Collocation Method (PCM) is employed to reproduce a large in silico dataset by approximating the G&R simulation outputs. A Deep Belief Network (DBN) is then trained to provide fast predictions of patient-specific AAA expansion, using both in silico data and patients' follow-up data. Follow-up Computer Tomography (CT) scan images from 20 patients are employed to demonstrate the effectiveness and the feasibility of the proposed model. The test results show that the DBN is able to predict the enlargements of AAAs with an average relative error of 3.1%, which outperforms the classical mixed-effect model by 65%.

Data-driven cardiovascular flow modelling: examples and opportunities

“…Neural networks are a powerful class of machine learning techniques and may be thought of as an approximator to complex nonlinear functions. A popular application in cardiovascular research is learning clinical outcomes based on different input parameters derived from patient-specific simulations [117,118]. We may think of this as an advanced multiparameter regression framework where we have several input parameters, different clinical outcomes, and large databases where we are interested in learning/fitting a statistical relationship between the data, which is achieved through the learning process.…”

Virtual Surgical Planning: Modeling from the Present to the Future

“… A model-based method can be used to simulate and fill in the gaps in the database used for data-based prediction method. Jiang et al [ 36 ] and Vavourakis et al [ 33 ] utilized such techniques for predicting changes in abdominal aortic aneurysms and breast tissue deformation, respectively. Data-driven techniques can be used to create population-specific parameters required by model-based methods [ 37 ].…”

Data-driven cardiovascular flow modeling: examples and opportunities

“…Machine learning and neural networks Neural networks are a powerful class of machine learning techniques and may be thought of as an approximator to complex nonlinear functions. A popular application in cardiovascular research is learning clinical outcomes based on different input parameters derived from patient-specific simulations [107,108]. We may think of this as an advanced multi-parameter regression framework where we have several input parameters, different clinical outcomes, and large databases where we are interested in learning/fitting a statistical relationship between the data, which is achieved through the learning process.…”