This paper presents an approach to automatic recognition of functional states of cellular groups based on data obtained by recording the spectrum of the brain acoustic signals. Data collection was performed using a prototype of the medical device ‘Brain Acoustic Field Spectrum Recorder RS AEG-01’ in a medical institution for patients with confirmed diagnoses. After 160 s of frame recording, the device outputs two matrices of 4200 floating-point numbers each, both being obtained through spectral analysis and accumulation of acoustic signals in the range from 0.13 to 27 Hz. Functional state classification was performed using a 1D convolutional neural network, with its architecture and hyperparameters being selected by an evolutionary algorithm within a computing cluster. The distinguishable classes included: practically healthy individuals, amenorrhea, operated cancer, breast cancer, and endometriosis. The Classification accuracy reached 0.93 for the test sample.