UC San Diego

Assessing regional cardiac function has many diagnostic implications in the treatment of various cardiac diseases such as myocardial ischemia, cardiac dyssynchrony, etc. Unfortunately, there is no accepted gold standard for the quantitative measurement of regional cardiac function. Current methods to assess regional cardiac function primarily rely on the subjective visual analysis of cardiac images.

The method described in this thesis processes high resolution cardiothoracic CT scans to segment out and characterize the left ventricular endocardium. We hypothesize that the fractal dimension is an effective parameter to characterize and quantify the regional roughness/structure of the endocardium, which changes across the cardiac cycle.

The method was tested on 10 normal cases and on one case with heart failure. All 10 normal cases showed a consistent and uniform decrease in the fractal dimension values from end diastole to end systole whereas the heart failure case showed no considerable change, differing by 3 standard deviations from the normal cohort. The method was consistent in distinguishing normal from abnormal for all possible sizes of the local neighborhood, but showing variation in the absolute values of the fractal characterization.

Within a definite range, the method was not significantly sensitive (< 1% variation) to the initial threshold values input by the user. A study with three independent volunteers showed that the user was most likely to select a threshold value within this insensitive range.

The method developed is non-invasive, simple to use with minimum operator involvement, and computationally inexpensive.