Ice particles in clouds are a known threat to commercial aviation. They can form areas with high ice water content at cruise altitudes where severe icing conditions may be sustained and not detected with current on-board technologies. In this article, a comparative study between two ice particle mass retrieval methods is presented. Both methods use in-situ measurements of the Particle Size Distribution (PSD) and Ice Water Content (IWC) as input to establish mass-diameter relationships (henceforth m-D). The first method follows the classical approach where the form of the m-D relationship is limited to power laws. The second method is based on an inverse problem approach: it leverages optimal use of in-situ measurements to waive the power law constraint. Both methods are tested on an in-situ measurement dataset collected during one research flight in an oceanic Mesoscale Convective System (MCS) with measured IWC values frequently exceeding 1.0 g.m-3. This flight was carried out in Darwin (Australia) area during the High Altitude Ice Crystals/High Ice Water Content 2014 international field campaign. Results show significant differences in crystal masses retrieved from both methods. The first outcome of the comparison is that IWC values calculated from the non-classical method are significantly closer to the measured IWC values than those calculated from the classical approach (the overall error is decreased by almost a factor 2). Another advantage of the non-classical approach is the increased latitude in the retrieval process to actually capture some size dependent microphysical processes. Preliminary results reveal different regimes in the variation of particle effective density with size. These variations are not captured by the classical mass retrieval methods where the m-D power law assumption restricts the shape of effective density versus size to a strictly decreasing polynomial curve. The Median Mass Diameter values (MMD) retrieved along flight path with both methods are found to differ: MMDs calculated from the inverse problem approach are consistently 20% smaller than those produced by the classical power law.