Lecture 16 + 17- Modelling Flashcards
What is the primary aim of a diagnostic model?
To quantify a characteristic of a system that determines the current health or disease.
What is the primary aim of a prognostic model?
To predict how health or disease might evolve in the future.
How should model complexity be determined?
The model should be no more complex than necessary otherwise waste resources, difficult to parameterise etc
BUT complex enough to capture key phenomena to answer specific questions
Steps of patient specific modelling?
- Acquisition of the geometry (scanning, segmentation etc)
- Computation analysis
- Data post processing
What is an idealised model?
Simplified representation which capture key features
What is image guided modelling?
Integration of imaging with computational modelling
What are the 4 main reasons for the increase in patient specific modelling since 2000?
Improvements in computer power (Moore’s law)
Improvements in modelling software (ANSYS, Abaqus etc.)
Availability of high-resolution medical imaging (MRI etc.)
Potential of biomechanical measurements for clinical decision-making and surgical planning
What features would the ideal imaging modality possess?
High resolution, low noise, artefact free, high contrast between tissues
Which imaging modality comes closest to the ideal? Why?
CT (Computed Tomography) - It can distinguish the thrombus from the lumen in aneurysms.
Main problems with CT for aneurysmal modelling?
It cannot measure wall thickness due to resolution limitations and insufficient contrast between tissues
In atherosclerosis, the resolution cannot show detailed structural differences in the plaque
Main advantages of MRI?
Image contrast is higher to distinguish between tissues, and can enable some visualisation of aortic wall
It enables visualisation of atherosclerotic plaque components
Main disadvantages of MRI?
Data acquisition times are long, patients must hold their breath to image, but this comes with significant noise
XY resolution is great, but slice thickness is much greater than CT and so Z resolution is poor. This makes it very difficult to measure WSS
Main advantages of ultrasound?
Ultrasound data may be acquired using externally produced transducers
State of the art ultrasound gives excellent resolution for PSM
Main disadvantages of ultrasound?
There are problems in practise with registration, low resolution, and loss of data due to to calcifications and bowel gas
How can you mitigate the limitations of ultrasound?
By using intravascular ultrasound (higher frequency) where the transducer is placed inside the vessel
What is optical coherence tomography (OCT)?
Use of coherent light to capture micro-resolution (it is effectively optic ultrasound)
Main advantages and disadvantages of OCT?
Extremely high resolution (higher than any other modality), and instant imaging
But it is an invasive technique
What is segmentation?
Identification of surfaces of interest in the organ
It can be manual, automated or semi-automated
What must happen after segmentation of a surface and why?
Geometry must be further prepared to remove artefact and smooth the surface, so that it is reflective of reality
What happens in modelling after the geometry is obtained?
Mesh generation, Computational modelling, and post processing.
What are some problems associated with the increasing use of AI and machine learning in modelling?
AI/ML produces outcomes based on previous inputs from a data set. Therefore acts as a “black box” and does not have explanatory power based on physical understanding.
What are the pros of a 0D model?
can be used to describe system-level response and interaction with different organs
directly comparable with clinical measurements
no need of expensive computing resources
What are the cons of a 0D model?
frequency-domain representations neglects convective acceleration term
not capable of representing pulse wave propagation, WSS/pressure
distributions, etc.
some model parameters difficult to identify, easily affecting accuracy of results
