In this study we show how the consistency conditions of the Radon transform can be used to aid attenuation correction in PET using short transmission scans. The technique is expected to be useful in situations where limited time is available for transmission imaging (e.g. wholebody PET). The proposed method uses two-minute transmission scans that are reconstructed and then segmented into regions of uniform attenuation. Consistency information is used to determine the thresholds for segmentation of the transmission image and the attenuation coefficients. In particular a downhill simplex algorithm is used to find the parameters that are most consistent with the measured emission data. The main advantage of the proposed technique over conventional segmentation methods (which work purely on the transmission data) is that the emission data are used to drive the segmentation process. Therefore, the technique should produce the attenuation image that is most appropriate for attenuation correction. The algorithm is tested using simulated data and compared with an adaptive thresholding technique using clinical PET data. The results show that the method produces reconstructed images with similar accuracy and noise levels to those obtained using a 10 minute transmission scan. The results also show that the proposed technique has some advantages over the adaptive thresholding technique when imaging the abdomen. We conclude that the proposed technique may provide a viable means of producing quantitatively accurate whole-body PET images in a clinically feasible time.