Electromagnetic induction tomography (EMT) is an emerging tomography technique which utilizes inductive sensors to image the conductivity distribution of an object. This paper introduces a newly established EMT system with 32 sensors, which is specifically designed to study the effect of missing data on the quality of reconstructed images in EMT. Missing data are investigated by systematically removing the coil sensors through the undersampling process and limited angle imaging. The EMT system with 32 sensors provides a data set consisting of 496 measurements, where some of the data might be missing due to the nature of imaging objectives. To examine a range of missing data sets, two experimental scenarios are completed: undersampling measurements and limited angle imaging. The former is carried out by evenly undersampling 4, 8 and 16 sensors from a 32-sensor coil array and the latter is investigated by using limited angles of 45°, 90°, 180° and 270°, compared to 360° full angle imaging. An edge FEM is used to calculate the forward problem and a linear algorithm is implemented as an inverse solver to reconstruct images. An image quality measure and 1D graph of conductivity distribution are adopted to quantify the effect of missing data on EMT images through experimental evaluation.