This paper presents a novel stochastic approach for criticality assessment in advanced driver assistance systems (ADAS). Modern assistance system rely on multiple information sources (e.g. radars, image processing) which provide data with a relative accuracy. As a consequence, criticality assessment for future ADAS tend to use stochastic methods instead of deterministic ones in order to consider such uncertainties. Our new method estimates the collision probability and also the Time-To-Collision (TTC) probability distribution for more robust and real-time decision making. The presented method is able to handle complex traffic situations with any number of traffic participants and abritrary trajectories.