Real-Time Virtual Pathology Using Signal Analysis and Synthesis

This dissertation discusses the modeling and simulation (M& S) research in the area of real-time virtual pathology using signal analysis and synthesis. The goal of this research is to contribute to the research in the M&S area of generating simulated outputs of medical diagnostics tools to supplement training of medical students with human patient role players. To become clinically competent physicians, medical students must become skilled in the areas of doctor-patient communication, eliciting the patient\u27s history, and performing the physical exam. The use of Standardized Patients (SPs), individuals trained to realistically portray patients, has become common practice. SPs provide the medical student with a means to learn in a safe, realistic setting, while providing a way to reliably test students\u27 clinical skills. The range of clinical problems an SP can portray, however, is limited. SPs are usually healthy individuals with few or no abnormal physical findings. Some SPs have been trained to simulate physical abnormalities, such as breathing through one lung, voluntarily and increasing blood pressure. But, there are many abnormalities that SPs cannot simulate. The research encompassed developing methods and algorithms to be incorporated into the previous work of McKenzie, el al. [1]–[3] for simulating abnormal heart sounds in a Standardized Patient (SP), which may be utilized in a modified electronic stethoscope. The methods and algorithms are specific to the real-time modeling of human body sounds through modifying the sounds from a real person with various abnormalities. The main focus of the research involved applying methods from tempo and beat analysis of acoustic musical signals for heart signal analysis, specifically in detecting the heart rate and heartbeat locations. In addition, the research included an investigation and selection of an adaptive noise cancellation filtering method to separate heart sounds from lung sounds. A model was developed to use a heart/lung sound signal as input to efficiently and accurately separate heart sound and lung sound signals, characterize the heart sound signal when appropriate, replace the heart or lung sound signal with a reference pathology signal containing an abnormality such as a crackle or murmur, and then recombine the original heart or lung sound signal with the modified pathology signal for presentation to the student. After completion of the development of the model, the model was validated. The validation included both a qualitative assessment and a quantitative assessment. The qualitative assessment drew on the visual and auditory analysis of SMEs, and the quantitative assessment utilized simulated data to verify key portions of the model