Charles Metz and colleagues from the Department at Radiology, University of Chicago have created a computer program which does Maximum Likelihood Estimates of ROC data. A web interface for ROCFIT is available.
Version 9 and above contain an ROC graph module. ROC analyses can be computed for non-parametric and bi-negative exponential models. The normal distribution model is not offered.
Although this program does not have a signal detection module, per se, it does have built-in statistics functions convenient for working with regression (TREND, SLOPE, and INTERCEPT) and converting p and z values (NORMSDIST, and NORMSINV).