Computational Models for Estimating Liver Iron Overload with the Magnetic Iron Detector

Abstract—An accurate measurement of the liver iron over-load is essential for the management of diseases such as tha-lassemia and hemocromathosis. The Magnetic Iron Detector is a susceptometer, which measures the total iron overload in the liver and has been used on more than 800 patients of Galliera Hospital (Genoa, Italy) since February 2005. The iron overload is obtained by calculating the difference between the measured magnetization signal and the patient’s background signal, which is the magnetization signal that would be measured for that patient with a normal iron content. This study presents two models for calculating the background signal using the measurements and the anthropometric features of 84 healthy volunteers. The first model introduces a statistical correction to the signal computed from the body shape of the subject assuming it to be made of water. The second model is based on statistical learning and learns from the volunteers ’ data a mapping from the anthropometric features to the background signal. We also present two approaches to combine the models. The assessment of the models on the 84 volunteers show that the performances of the models are comparable and that we can confidently estimate the background signals of patients. The model sensitivity (0.9 g) allows the physicians to monitor the iron overload variations due to the therapy. These models are currently in use at the Galliera Hospital