Fast detection of emerging diseases is important for containing their spread and treating patients effectively. When using medical imaging for detection, local anomalies are relevant, but often novel diseases involve familiar disease patterns in new spatial distributions. Therefore, established local anomaly detection approaches may fail to identify them as new. Here, we present a novel approach to detect the emergence of new disease phenotypes exhibiting distinct spatial distribution patterns of lesions. We first identify anomalies in lung CT data, and then compare their distribution in continually acquired new patient cohorts with a historic patient population observed over a long prior period. We evaluate how accumulated evidence collected in the stream of patients enables detection of an emerging disease onset. Results show that in a gram-matrix based representation derived from intermediate layers of a three-dimensional convolutional neural network, newly emerging clusters indicate emerging diseases.