The analysis of spatiotemporal patterns can provide clues about disease spread by assessing if the spatial pattern of diseased plants at one date is associated with the pattern of previously diseased plants. No generic statistical test was available to answer this question for spatiotemporal maps of binary data (healthy or diseased plants) in regular plantings (e.g., orchards). Here we describe a Monte Carlo test of the hypothesis that the location of newly diseased plants is independent of the location of previously diseased plants, even when the disease is spatially aggregated within each assessment period. This spatiotemporal test is designed to cope with the censorship arising on a lattice when plants are missing or cannot recover between the two dates. Expected patterns are simulated by shifting on a torus the whole pattern at the second date relatively to the pattern at the first date. For each simulation, we discard the censored points from observed and simulated data. In case of a positive association between disease patterns at two dates, the distances between newly and previously diseased trees should be smaller in the observed than in the simulated patterns. As an illustration, we analysed the dependence between patterns of trees showing Plum pox virus symptoms at two dates.