Radiation portal monitors positioned at seaports form a crucial part of national security infrastructure by providing a means to deter and detect the illicit transportation of nuclear and radioactive materials. With around 90% of global goods transported by sea, it is essential that international freight can be rapidly and non-invasively screened for materials that have potential use in radiological dispersal devices or improvised nuclear explosive devices. To maximise detection efficiency, conventional radiation portal monitors use low-cost, large-area scintillator detectors, usually polyvinyl toluene plastic, that maximises the collection of gamma radiation as any concealed sources pass through the portal monitor. While advantageous in many regards, a drawback of using large area scintillators in radiation portal monitors is that they are non-directional; it is impossible to determine the direction from which the measured radiation originates. In this work, an algebraic method called maximum likelihood expectation maximisation (MLEM) reconstruction is applied to simulated data from the non-directional detectors found in conventional radiation portal monitors. The MLEM method is well studied in the field of medical imaging where high resolution, small, and directional detectors are used and has more recently been applied to radiation mapping problems with non-directional detectors. To use the MLEM reconstruction method in shipping container screening the response function of each detector in the radiation portal monitor to a radioactive pointsource is first characterised on a regular grid of points in the centre of the portal monitor. Through making assumptions about the cargo, such as homogeneity throughout the shipping container and prior knowledge of the material composition, an attempt is made to include the attenuating effect of the cargo in the detector response function. With the detector response function defined for a given cargo type, three variants of the MLEM method are used to characterise concealed point-sources in the cargo for reference cargo types, including a container filled with scrap metal. Preliminary results show the algorithm is capable of localising the source to within a few voxels in the discretised space in the container and estimating the activity to the correct order of magnitude.