Evaluation of Liver Function in Healthy Subjects and Liver Disease Patients Using BOLD MRI

Abstract

The liver is a multi-function organ that plays important roles in nutrient metabolism, biochemical transformations and blood detoxification. The purpose of the current work was to optimize Blood Oxygen Level Dependent (BOLD) liver functional MR imaging and analysis to allow the distinction between healthy volunteers and subjects with chronic liver disorders known to lead to fibrosis and reduced liver function (in this case, Hepatitis-C). Liver BOLD signal can be modulated by breathing 100% 0 2 or through intake of a meal. Previous results using these stimuli have been inconclusive when comparing healthy and diseased livers. In addition, liver BOLD analysis has been traditionally carried out using general linear models (GLM). Since the liver has a dual blood supply (portal and arterial derived), its resultant haemodynamic response is complex. This makes it too difficult to employ GLM approaches, as they require the prediction and modeling of a response function. We chose a model-free, or data-driven approach, called principle component analysis (PCA) to analyze liver data. Initial optimization was done by determining the time of maximal hepatic portal vein (HPV) blood flow following ingestion of a controlled meal (235 mL of Ensure Plus®). Statistically significant increases in HPV flow resulted at all measurement intervals, with the maximal postprandial change (71% increase in comparison to the baseline flow) at thirty minutes after ingestion. Implementing acquisition and analysis optimizations with our dual liver challenge model (hyperoxia cycling in pre- and postprandial states), the PCA approach was able to detect all of the diseased livers (n=6), while missing four of the healthy subjects (n=ll). The GLM technique, on the other hand, did not detect two of the patients and two of the healthy subjects. Thus, if this liver challenge is to be used as a screening tool, a model-free data analysis approach is suggested as more appropriate since it minimizes the chances of reporting false-negative results (based on this preliminary cohort). Although more false positives were detected with this method, it is of less concern seeing as these inaccuracies can be screened using simple blood tests. Promising results were obtained in this project, however, further studies using data-driven approaches such as partial least squares (PLS) are needed.