The invention discloses a CSI index extraction method for an anesthesia depth monitor. An electroencephalogram collection circuit is provided with an input end and an output end; a weak electroencephalogram is converted into a size which can be collected by an AD chip and then transmitted to a microprocessor; the microprocessor obtains 4 major parameters (alpha)ratio, (beta)ratio, (beta)ratio-(alpha)ratio and BS% about CSI through the processes of filtering, fast discrete Fourier transform, frequency band energy calculation and the like; and based on the 4 major parameters, the anesthesia depth index CSI is finally obtained through an adaptive neuro-fuzzy inference system. In the invention, the CSI index is calculated by combining the fast discrete Fourier transform and the adaptive neuro-fuzzy inference system, and other nonlinear interference is avoided; and moreover, the obtained CSI index is displayed on a display screen to provide a reliable basis for timely and accurate anesthesia operation of a patient by medical personnel.