Identification, implementation, and validation of advanced metrics for automated aorta characterization ■
Subject ■
Aortic pathologies such as aneurysms and dissections [1] remain critical conditions requiring accurate imaging-based assessment for diagnosis, risk stratification, and treatment planning. In clinical routine, these conditions are primarily evaluated using CT and MRI, where quantitative measurements play a central role in guiding clinical decisions. Current clinical practice relies on a set of standardized geometric measurements, as defined in the 2022 ACC/AHA guidelines [2], including the maximal aortic diameter and diameters at predefined anatomical landmarks. Traditionally, these measurements are performed manually and are considered the clinical reference. However, manual assessment is time-consuming and subject to inter-observer variability, which may affect reproducibility and consistency in follow-up.
Recent advances in medical image analysis, especially deep learning, have enabled the automation of these standard measurements. These approaches have demonstrated accuracy comparable to expert annotations while significantly reducing analysis time and improving reproducibility. More specifically, our group has introduced the AORta Automated Measurement (AORAM) tool capable of such measurements [3]. Despite these developments, standard clinical metrics provide only a limited description of the aorta, reducing a complex three-dimensional structure to a few local measurements. Growing evidence in the literature suggests that aortic disease is influenced by global morphology and geometric variability that are not captured by diameter alone. For example, aortic arch morphology has been associated with disease development and risk [4], while metrics such as aortic root acircularity quantify deviations from ideal cross-sectional geometry and capture subtle structural changes [5]. These observations motivate extending current aortic characterization frameworks beyond standard clinical measurements. Building on an existing automated pipeline for extracting guideline-based metrics, this project aims to integrate advanced morphological descriptors that capture both local and global properties of the aorta.
Recent advances in medical image analysis, especially deep learning, have enabled the automation of these standard measurements. These approaches have demonstrated accuracy comparable to expert annotations while significantly reducing analysis time and improving reproducibility. More specifically, our group has introduced the AORta Automated Measurement (AORAM) tool capable of such measurements [3]. Despite these developments, standard clinical metrics provide only a limited description of the aorta, reducing a complex three-dimensional structure to a few local measurements. Growing evidence in the literature suggests that aortic disease is influenced by global morphology and geometric variability that are not captured by diameter alone. For example, aortic arch morphology has been associated with disease development and risk [4], while metrics such as aortic root acircularity quantify deviations from ideal cross-sectional geometry and capture subtle structural changes [5]. These observations motivate extending current aortic characterization frameworks beyond standard clinical measurements. Building on an existing automated pipeline for extracting guideline-based metrics, this project aims to integrate advanced morphological descriptors that capture both local and global properties of the aorta.
Kind of work ■
The main objective is to identify, implement, and validate relevant metrics for aortic characterization through our AORta Automated Measurement tool (AORAM).
Specific objectives are to:
(1) identify aorta characterization metrics used in clinical practice and clinical/technical research
(2) implement the most relevant metrics and integrate them in AORAM
(3) validate the implemented aorta characterization metrics
(4) optimize metrics for efficiency and robustness.
Specific objectives are to:
(1) identify aorta characterization metrics used in clinical practice and clinical/technical research
(2) implement the most relevant metrics and integrate them in AORAM
(3) validate the implemented aorta characterization metrics
(4) optimize metrics for efficiency and robustness.
Framework of the Thesis ■
The project will consist of:
•
A literature review to identify standard metrics and relevant advanced metrics for aorta characterization.
•
An analysis of the available datasets that can be used for testing and validating the identified metrics.
•
Implementing the most relevant metrics for aorta characterization in Python and integrating them with the existing AORAM tool.
•
Validating metrics on relevant datasets and comparing automated metrics to manual ones.
•
Optimizing the metrics for efficiency and robustness.
•
A literature review to identify standard metrics and relevant advanced metrics for aorta characterization.
•
An analysis of the available datasets that can be used for testing and validating the identified metrics.
•
Implementing the most relevant metrics for aorta characterization in Python and integrating them with the existing AORAM tool.
•
Validating metrics on relevant datasets and comparing automated metrics to manual ones.
•
Optimizing the metrics for efficiency and robustness.