Classification of Aortic Shape with Topographical Pair Correlation Functions

Area of Science:

• Medical imaging analysis
• Computational anatomy
• Cardiovascular diagnostics

Background:

• Quantitative descriptors of medical images aid in diagnosing conditions like aortic dissections.
• Aortic shape analysis, particularly of the true and false lumens, is crucial for predicting disease state and outcomes of thoracic endovascular repair (TEVAR).

Purpose of the Study:

• To introduce a novel method, the topographical pair correlation function (TPCF), for quantifying aortic shape from medical images.
• To extract quantitative features from TPCF curves for improved disease classification and prediction of TEVAR success.

Main Methods:

• Developed a method to calculate the TPCF, describing spatial correlations of Gaussian curvature, mean curvature, shape index, and bending ratio on a meshed image surface.
• Applied TPCF as a metric for aortic shape analysis and extracted features from the resulting correlation curves.
• Utilized TPCF parameterized by shape index for classification tasks.

Main Results:

• The TPCF, when parameterized by shape index, achieved a 0.95 area under the curve for classifying disease presence and/or impending TEVAR success.
• Features extracted from the TPCF demonstrated superior performance compared to single-point statistics in classifying aortic disease states.
• The TPCF effectively captures structural correlations in anatomical data.

Conclusions:

• The TPCF is a powerful tool for quantitative shape analysis in medical imaging, particularly for the aorta.
• This method offers improved diagnostic capabilities for aortic dissections and prediction of treatment outcomes.
• TPCF analysis holds broad potential for characterizing structural correlations in diverse anatomical datasets.