Microthermistors are put on the surface of cerebral cortex to monitor local cerebral blood flow (CBF) continuously with minimal tissue damage and disturbance to the normal physiological state. Using a distributed, dynamic model of the measurement system, we simulated the effects of this flow measurement method under isothermal and adiabatic boundary conditions. Numerical results show that the adiabatic boundary condition can provide maximal sensitivity to perfusion changes at physiological perfusion levels. The constant power and constant temperature operating modes are compared in terms of output relation, sensitivity, and frequency response through analytical and numerical solutions. While the steady-state relations between thermistor measurements and perfusion for the two modes do not differ significantly, the constant temperature mode has better frequency response. Analytical results show that the relative sensitivity is the same for the two modes and is approximately proportional to the radius of thermistor. If there is an unperfused layer surrounding the thermistor, the sensitivity will decrease as the thickness of the layer increases. Simulations predict that the thermal measurement has a low-pass frequency response and the cutoff frequency is inversely proportional to the probe surface area. The results provide a theoretical foundation to the optimal design of thermistor probe for continuous CBF measurement from tissue surface.