Performance of a large-area, gamma imager using a point source injection technique

We recently developed a fieldable large-area, coded-aperture, gamma imager (the Large Area Imager - LAI). The instrument was developed to detect weak radiation sources in a fluctuating natural background. Ideally, the efficacy of the instrument is determined using a receiver-operator statistics generated from measurement data in terms of probability of detection versus probability of false alarm. However, due to the impracticality of hiding many sources in public areas, it is difficult to measure the data required to generate receiver-operator characteristic (ROC) curves. Instead, we developed a high statistics "model source" from measurements of a real point source. The model source is then injected into data collected from the world at large where, presumably, no source exists. In this paper we have applied this "source injection" technique to evaluate the performance of the gamma imager. We injected the model source at many locations parallel to the imager driving direction keeping the source to imager distance fixed. From this data, we generated probabilities of detection and false alarms and plotted ROC curves for 1, 1.25 and 1.5 mCi sources placed at 50-m from the imager. We also developed ROC curves for 1 mCi source injected at 35, 40, 45 and 50-m from the imager. The result shows that this prototype instrument provides excellent performance for a 1-mCi source injected at 50-m perpendicular distance from the imager in a single pass at an average velocity of 11 m/s.