[Glaucoma perimetry--optimization of test point patterns with an information index].

An experimental clinical study on 274 glaucomatous eyes was performed to investigate extensively the relation between test point patterns and information in glaucoma. We used the Humphrey Field Analyzer 620 and a special high-density pattern of 165 points containing all classic patterns of threshold perimetry (31, 32, 61). By deleting points the classic patterns, as well as other patterns, were deduced and compared. Using objective criteria, we measured the frequency of scotoma detection, modification, and refinement that occurred with pattern condensation in different areas. The average gain of information by new points, expressed by an information index, showed an exponential decrease with increasing spatial resolution. Also, the information index decreased toward the periphery for the same step. Optimal patterns, which have an equilibrated information index in all areas, should be denser in the center and coarser at the periphery. Two of these, a 3 degrees-4.2 degrees-6 degrees pattern (called "30-S") and a 4.2 degrees-6 degrees-8.5 degrees pattern ("half 30-S") were compared to the classic patterns and appeared to be superior, best demonstrable in the detection rate: each of the classic patterns (31, 32) missed 15% of the early glaucomatous defects, the adequate central condensed pattern only half of this. In the combination the classic patterns still missed 2.5% of the early defects; the corresponding central condensed pattern missed none. Moreover, these central condensed patterns have fewer points, so that test time is shorter. Changing to these patterns during follow-up is no problem, because this new family of test patterns is deduced from the classic patterns and corresponds in the majority of test locations.