Background Cavitation has been claimed partly responsible for the increased risk of thromboembolic complications, hemolysis, and fatal valve failure seen in mechanical heart valve patients. In vivo studies have investigated cavitation using high-pass filtering of the high-frequency pressure fluctuations with the root mean square values as an assessment of intensities. In vitro studies have shown that this well-known method may not be ideal owing to loss of data as a consequence of filtering, and because it requires a priori knowledge of the valve resonance pattern. Therefore, a new method has been developed, which decomposes the signal into nondeterministic (cavitation) and deterministic (valve resonance) signal components, and hence decreases data loss. This study aimed to evaluate cavitation in patients with mechanical, biological, and native heart valves both intraoperatively and postoperatively using the new method. Methods High-frequency pressure fluctuations were measured by a hydrophone intraoperatively and postoperatively in 14 patients with mechanical valves, 10 patients with normal aortic valves, and 5 patients with bioprosthesis. The total signal energy was evaluated as nondeterministic and deterministic energies. Results Nondeterministic energies were verified both intraoperatively and postoperatively in all patients who had a mechanical valve; this finding confirms the cavitation potential of mechanical valves. None of the data recorded in patients with bioprosthetic or native valves contained nondeterministic energy. Conclusions The study confirms the presence of cavitation in mechanical heart valve patients using the nondeterministic energy of high-frequency pressure fluctuations as a quantitative measure of cavitation both intraoperatively and postoperatively.