This article describes a method to improve the work efficiency of the software complex for automatic segmentation and analysis of lumbar spine MRI images in conditions of dynamic changes. For the modules of the software complex, an event tree and a state graph developed by authors, which will allow taking into account the conditions leading to software complex failures. These conditions with a set of certain factors at different time intervals can break down all modules of the software complex. The developed models can reduce the program errors, debugging time and recovery of a non-functioning system. A system of Kolmogorov-Chapman differential equations based on the state graph. For the system of differential equations, a utility software developed via the Python programming language and SciPy library, which allows estimating the level of probability of failure of system modules over a certain time. The behavior of the software complex modelled under various initial conditions. As result of the solution of the differential equations system, the influence of the intensities of recovery and failure of elements on the operability of the system modules investigated. For the various initial states of the system, an analysis, based on which recommendation rules created for the management personnel of the software complex, which will avoid a critical combination of events leading to a system failure.