Background ― The treatment of the cancer, especially in more aggressive metastatic forms is more effective at early disease stage. However, existing diagnostic techniques are not sensitive enough for early cancer detection. An alternative, perspective diagnostic approach can be based on photoacoustic (PA) method of irradiation of cancer cells in biotissue, blood and lymph by laser pulses. The fast thermal expansion of heated zones into cells associated with intrinsic or artificial PA contrast agents leads to generation of acoustic waves detected with ultrasound transducers. In particular, melanoma cells with melanin as a PA marker are darker than normal red blood cells and, therefore, produce greater acoustic responses. This technique can theoretically detect even a single cancer cell in the tissue and blood background; however, a robust algorithm of automated response detection is yet to be developed. Objective ― The main aim is to develop the approach for data pre-analysis that can improve the sensitivity and noise resistance of the automated in individual cancer cell detection algorithm, based on estimation of the amplitude of the acoustic responses. Methods ― Acoustic responses were obtained from a round polyurethane tube with human blood, or solution of the mouse melanoma cells in 10 mol/L concentration. In control experiments the laser was blocked by an opaque film. Many (up to 1000) acoustic responses were obtained from normal blood cells and pigmented cancer cells. Spectral analysis of the acoustic responses was used to find the spectral ranges that provide valuable diagnostic information with the sufficient signal-to-noise ratio. Results ― It was estimated that relevant diagnostics information in the acoustic responses is limited to the 0-12 MHz frequency band. Application of the 8th order low-pass Butterwort filter with 12 MHz cut-off frequency improved the signal-to-noise ratio from 21.14±10.39 to 110.81±56.94 for the cancer-related responses, and from 1.04±0.1 to 2.23±0.33 for the normal blood responses. Conclusions ― Adoption of low-pass filtering during the pre-analysis of acoustic responses results in better sensitivity of automated cancer cells detection algorithm.