AI techniques for computational cardiology

Abstract

Biomedical research is at a critical point at present. The research has led to an enormous amount of data and models describing these data, but approaches for application, formalization and integration of this knowledge from the molecular to the system level are still topics of ongoing research and certainly far from fully developed.

Also in cardiology the different anatomical and physiological constituents as well as the coupling between them are being researched in increasing detail and are often described using computer-based models. But for this domain an integrative framework is still missing. [1]

An artificial neural network is a system based on the operation of biological neural networks, in other words, is an emulation of biological neural system. Although computing these days is truly advanced, there are certain tasks that a program made for a common microprocessor is unable to perform; even so a software implementation of a neural network can be made with their advantages and disadvantages. [2]

The term ‘artificial intelligence’ was coined in 1956 by Professor John McCarthy at the Massachussetts Institute of Technology (MIT).While the ambitious goal of ‘making computers think and behave like humans’ is still largely unrealized (to date no computer has ever passed the Turing test—defined as the capability to engage undetected in a natural language conversation with a human party), human behavior has been rather effectively emulated by computers in game playing, as well as in select scientific and medical applications. For the purposes of this article, a software application will be considered to be ‘intelligent’ if it follows different courses of action based on the result of evaluation(s), e.g. expressed by ‘if then’ statements in the code. By this criterion, an algorithm that filters an image using a filter of predefined characteristics would not qualify as intelligent, whereas an algorithm that tailors the amount and type of filtration to image quality and count statistics would. [8]

Problem Description

An important application of modeling in biomedical research is to understand mechanisms of heart failure, which is a leading cause of death in humans. Usage of appropriate models simplifies development and validation of drugs and medical devices.

Modeling allows the exploration of a product’s side effects, which is of particular importance for the product’s approval.

Cardiac models provide a simplified description of the heart and can exist in a physical and mathematical representation. Mathematical models are commonly computer-based and applied in numerical simulations. Modeling of the heart is subject of manifold, often interdisciplinary research activities undertaken in academic and commercial fields. The activities range from the description of molecular structures and interactions to reconstruction of gross anatomy and whole heart electro-mechanic behavior. Explorations of the heart were performed as early as in the middle ages and the renaissance. First models of cardiac anatomy and physiology were constructed, which are now considered as outdated resulting from new insights.

In recent decades new insights in the function and structure of the heart have been gained by the availability of new measurement techniques. The molecular structure and arrangement of cardiac cells have been explored e.g. with new imaging techniques.

The increase in computing capability in recent years has simplified and speeded up significantly the realization of simulations. On the one hand, these simulations offer the possibility of reconstruction of previously performed measurements. On the other hand, apriori unknown behavior can be predicted and complex phenomena can be studied.

The application of modeling as a research, development and clinical tool is stressed as a promising attempt to address many problems in cardiology, heart surgery and biomedical engineering. The modeling includes the areas of anatomy, electrophysiology, excitation

propagation, force development and mechanics as well as the coupling of these areas.

Special attention is given to macroscopic and integrative modeling with the aim of reconstruction of whole heart behavior. At macroscopic level anatomical models provide a basis for electrophysiological and mechanical models. Of particular interest is the assessment of influence of cardiac deformation to initiation and propagation of electrical excitation and to the force development. Variant simulation studies are demonstrated, which provide information about this assessment.

Detailed modeling of the macroscopic cardiac anatomy is commonly performed on base of medical imaging systems, which are used in clinical routine and research. The resulting data is transformed with methods of digital image processing to obtain a representation of anatomy, which is suitable for the target application.

Digital image processing is the use of computer algoritms to perform image processing on digital images. [3]

Different levels of spatial description can be distinguished ranging from analytical, comprehensive approaches to detailed descriptions on base of millions of volume elements. In the next sections exemplary analytical models are described, an introduction to the modeling sources is given and some models created by digital image processing are demonstrated.

Commonly, ultrasonic (US), magnetic resonance (MRT), and computed tomography (CT) are used for the imaging of the heart. These imaging systems use the tissue dependent variations of the acoustic impedance leading to reflection of ultrasonic waves and of their scattering, of the absorption of X-rays, and of the resonance behavior of nuclei to get information of the tissue distribution inside a body.

An alternative data source for the macroscopic modeling of human anatomy provides the Visible Human Project of the National Library of Medicine, Bethesda, Maryland (USA). The ongoing project started 1990 and aims at the computerized representation of human bodies, which should be made available as standard for the medical and technical research. In a first approach selected corpses were imaged by CT and MRT as well as by specific photographic techniques. The selection of corpses was made by different criteria,

e.g. normal height and weight, no pathologic changes in anatomy, and age between 20 and 60.