SIGNAL PROCESSING

Author: MARTÍNEZ-CABEZA DE VACA ALAJARÍN JUAN DE LA CRUZ.
Year: 2005.
University: POLITÉCNICA DE CARTAGENA [www.upct.es].
Place of defense: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA DE TELECOMUNICACIÓN.
Place of preparation: ESCUELA TÉCNICA SUPERIOR DE INGENIERÍA DE TELECOMUNICACIÓN.

Summary: Over the past 30 years, the cardiac auscultation has been replaced by modern imaging techniques for diagnosing cardiovascular status, although still widely used as a screening technique. Currently, research in this field aims to develop automated systems that can help the physician to diagnose the state of valvular heart and prioritize waiting lists depending on the status of the patient. In this Doctoral Thesis has developed a comprehensive system of aid to remote diagnosis of cardiovascular disease based on the analysis of heart sounds generated by the heart. This system consists of two fundamental parts: one processing step that analyzes the fonocardiograma (FC) and determines the cardiovascular status, and efficient algorithms that compress the FCG for storage and transmission of remotely. It has also made a proposal for electronic architecture of a prototype that integrates these stethoscope intelligent processing and compression algorithms. To register signals to be used in the analysis has developed an environment for the specific purchase FCG, in order to provide the physician with the extent possible the task of data collection and recording a series of specific actions ( changing area of auscultation, conducting maneuvers heart, etc.) that need to be considered for the analysis and diagnosis is carried out correctly. For the analysis of FCG has developed a methodology that provides an overall structure for processing from the acquisition of FCG to obtain the final diagnosis. This methodology has two characteristics: it is modular, thus ensuring a degree of independence between the blocks that form processing, and hierarchical, so that the blocks are arranged in processing levels, making an abstraction gradually gives cough as that amounts to these levels. It has developed a set of algorithms that perform processing FCG for all the blocks that make up this hierarchy, from the acquisition until diagnosis, based upon analysis methodology. It has followed a pattern based on the detection of cardiac events in the FCG, similar to that of the doctor during auscultation. This has enabled design a system to help diagnose general purpose, which increases the extent of its use. For processing has been used only signal FCG without using signals assistants as electrocardiogram or pulse. While this fact has slightly increased the complexity of the analysis, on the other hand facilitates the design of the electronic device by failing to require additional sensors or merger of signals. Some of the processing blocks that form the hierarchy of analysis include: filtering, calculation of envelopes, calculating heart rhythm, detection and identification of events, feature extraction and pattern recognition. It is possible to extend the usefulness of the analysis system enfocándolo to telemedicine, to obtain remote diagnostics of cardiovascular status. For the efficient transmission of signals FCG has developed a compression method with specific losses (direct compression of FCG), based on wavelet transform. This method has been adapted to the characteristics of FCG, taken and thus able to obtain compression rates higher than those provided for the audio compression methods exist. It also has developed a method of selective compression FCG from genetic events detected (based understanding events), which provides even greater rate compression direct compression without diminishing the quality with which they viewed the events. The resulting 8 two obte aab nests processing show a high percentage of cardiac cycles in which all the events have been correctly detected and identified: 91.27% and 65.65% for searches without puffs and puffs, respectively. The classification of cardiac cycles into three categories (normal records, with breath holosistólico and breath mesosistólico) has provided success rates of 100%, 92.69% and 95.57%, respectively. As for the compression method based on events has obtained a rate compression between 20% and 70% higher than in direct compression to equal quality for cardiac events. These two methods have been compared to his time with the method OGG Vorbis audio compression, and between 2.6 and 2.6 times for compression-based events, depending on the quality with which they perform the compression. The processing and compression algorithms have been integrated into an application that allows their execution independently of the platform used for its development, in order to ensure maximum portability. This application (ASEPTIC) offers a graphical user interface that provides a user friendly user, as well as a mode of operation command line to provide information to other systems or monitoring that analysis at a higher level could solicitársela. Finally, it has made a proposal for electronic architecture of a prototype smart stethoscope, also defining the specifications to be taken. This prototype is based on the hierarchy of analysis developed, and integrates the processing and compression algorithms. The purpose of this device is to replace the traditional acoustic stethoscopes, and must be of a size, weight, and ease of use comparable to that of these, besides incorporating diagnostic capacity and high degree of autonomy. The result has been a set of electronic modules, which will be projected on a future reconfigurable electronic circuits.