Improving the Differential Diagnosis of Pneumonia and Congestive Heart Failure Us

DESCRIPTION (provided by applicant): Stethographics has developed automated lung and heart sound products, based on 3 granted U.S. patents and 2 FDA approvals. Our Pocket PC based system gathers sounds via a contact sensor in a simple and practical way. Th e system automatically detects and quantifies crackles, wheezes, rhonchi, squawks and heart murmurs. 3M Littmann, the world s largest stethoscope company, bundles all E4000 electronic stethoscopes with our Sound Analysis Software. We propose a research pl an that will lead to development of a smart stethoscope. In addition to extracting sound features like crackle count, wheeze rate, and heart murmur grade, the incorporated neural network algorithms will provide a probable cause of these abnormal sounds s uch as pneumonia, congestive heart failure, or heart abnormality. We expect the smart stethoscope to find its applications in many settings: in physician's offices, hospitals, nursing homes - essentially everywhere the stethoscope is used. In addition, new areas of exploitation include settings where doctoral expertise or stationary medical equipment is not always available, and nurse is the main source of medical help: on the ships, oil rigs, embassies and home care by visiting nurses. The diagnostic infor mation provided by the smart stethoscope can be used locally or telemetered. We have initiated this research by tackling two common illnesses: pneumonia (PN) and congestive heart failure (CHF). It is estimated that 5 million people in the United States hav e CHF. Although in many instances the diagnosis of these conditions is easily made, it is not uncommon, particularly in the Intensive Care Unit setting, for it to be unclear as to which illness a patient has. In cases of doubt the patient is often treated for both. Yet diuretics are likely not good for patients with pneumonia in the absence of coexisting heart failure and it is not good practice to subject patients to the risk of antibiotics unnecessarily. Our preliminary results in 151 patients with 2 or m ore crackles per breath (CHF=70; PN=81) indicate that the crackles differ significantly in these two conditions. Classification algorithms based on crackles features were able to separate the two disorders with a sensitivity of 0.91 and specificity of 0.82 . In Phase I we plan to retrospectively study the database of over 1,000 patients using pattern recognition methods in order to develop the expert system that can differentiate PN, CHF, interstitial pulmonary fibrosis (IPF), and normal patients. In Phase I I we will expand the system to include diagnosis of asthma, COPD, and cardiac murmurs. In Phase III we will incorporate the expert system into a smart stethoscope. PUBLIC HEALTH RELEVANCE: This research is expected to provide new medical diagnostic softwar e that can be incorporated into a smart stethoscope. The use of the smart stethoscope will be particularly relevant in settings where doctoral expertise or stationary medical equipment is not always available and nurse is the main source of medical help. A utomated diagnostics with the smart stethoscope can simplify and improve care for patients in nursing homes, especially by detecting early signs of pneumonia and home monitoring of patients with cardiopulmonary disorders.