Smart stethoscopes

A method of processing an acoustic heart signal, the method comprising:

• capturing an acoustic heart signal from a patient;
• processing the acoustic heart signal to provide an acoustic data sequence comprising a sequence of data elements representing the acoustic heart signal;
• classifying the sequence of data elements using a neural network to provide a sequence of acoustic data category classifications and associated probabilities, each acoustic data category classification classifying one or more of the data elements into one of a plurality of acoustic heart data categories with an associated probability;
• processing the sequence of acoustic data category classifications and associated probabilities using one or more sequence models, where each sequence model represents the likelihood of a sequence of heart sounds represented by the acoustic data category classifications and associated probabilities, wherein the processing identifies a most likely sequence of the heart sounds;
• and
• outputting data identifying the most likely sequence of the heart sounds.

A method of processing acoustic heart signal data, the method comprising:

• capturing time series acoustic heart signal data from a patient, the time series acoustic heart signal data representing the sound of the patient's heart;
• extracting acoustic features from the time series acoustic heart signal data to provide time series acoustic feature data;
• classifying the time series acoustic feature data using each of at least two heart models, a first heart model representing a normal heart and a second heart model represents a heart with a heart murmur;
• and
• determining which of the first model and the second model best fits the acoustic heart signal data;
• and
• outputting data indicating whether or not the patient's heart has a heart murmur when the second heart model fits the acoustic heart signal data better than the first model.

An electronic stethoscope system, the system comprising:

• an acoustic signal capture device to capture time series acoustic heart signal data from a patient;
• a neural network to classify the time series acoustic heart signal data into a plurality of heart sound categories to provide time series sound category data, wherein the time series sound category data comprises, for each of a succession of time intervals, category probability data representing a probability of the acoustic heart signal for the time interval falling into each of the heart sound categories;
• one or more heart state models each having a sequence of heart states;
• a system to fit the time series sound category data to the one or more heart state models and determine timing data for the sequence of heart states and a confidence value for the model fit;
• and
• an output to output one or both of a model fit indication dependent upon the confidence value and an indication of the timing data.