Optimization of thresholds for eating detection
First Claim
Patent Images
1. A method comprising:
- receiving, by a processor, a fundic impedance signal of a subject sensed during an initial calibration period using at least one implantable fundic sensor of an eating detection device;
receiving, by the processor, an indication of actual ingestion events of the subject during the initial calibration period;
performing, by the processor, an analysis of the fundic impedance signal and the actual ingestion events, using a cost function to find each of a plurality of cost function values for each of respective candidate thresholds by;
identifying, using an eating detection algorithm, detected ingestion events during the initial calibration period by comparing the fundic impedance signal to the candidate threshold over the initial calibration period,comparing the detected ingestion events to the actual ingestion events to find a number of the detected ingestion events that represent true positives, a number of the detected ingestion events that represent false positives, and a number of the actual ingestion events that represent false negatives, andcalculating the cost function value responsively to a measure of sensitivity of the eating detection algorithm and a measure of positive predictive value of the eating detection algorithm, wherein the sensitivity equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false negatives, and wherein the positive predictive value equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false positives;
setting, by the processor, a threshold by selecting one of the candidate thresholds as the threshold responsive to the cost function values, the crossing of which threshold by the fundic impedance signal is indicative of ingestion; and
storing, by the processor, the threshold in a memory of the eating detection device.
2 Assignments
0 Petitions
Accused Products
Abstract
A method is provided that includes providing a fundic impedance signal of a subject sensed during an initial calibration period using at least one implantable fundic sensor of an eating detection device. An indication is received of actual ingestion events of the subject during the initial calibration period. Using a cost function, an analysis is performed of the fundic impedance signal and the actual ingestion events. Responsively to the analysis, a threshold is set, the crossing of which by the fundic impedance signal is indicative of ingestion. The threshold is stored in a memory of the eating detection device. Other embodiments are also described.
-
Citations
28 Claims
-
1. A method comprising:
-
receiving, by a processor, a fundic impedance signal of a subject sensed during an initial calibration period using at least one implantable fundic sensor of an eating detection device; receiving, by the processor, an indication of actual ingestion events of the subject during the initial calibration period; performing, by the processor, an analysis of the fundic impedance signal and the actual ingestion events, using a cost function to find each of a plurality of cost function values for each of respective candidate thresholds by; identifying, using an eating detection algorithm, detected ingestion events during the initial calibration period by comparing the fundic impedance signal to the candidate threshold over the initial calibration period, comparing the detected ingestion events to the actual ingestion events to find a number of the detected ingestion events that represent true positives, a number of the detected ingestion events that represent false positives, and a number of the actual ingestion events that represent false negatives, and calculating the cost function value responsively to a measure of sensitivity of the eating detection algorithm and a measure of positive predictive value of the eating detection algorithm, wherein the sensitivity equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false negatives, and wherein the positive predictive value equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false positives; setting, by the processor, a threshold by selecting one of the candidate thresholds as the threshold responsive to the cost function values, the crossing of which threshold by the fundic impedance signal is indicative of ingestion; and storing, by the processor, the threshold in a memory of the eating detection device. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
-
14. Apparatus for use with an eating detection device having a memory and at least one implantable fundic sensor, the apparatus comprising:
-
an interface, configured to communicate with the eating detecting device; and a processor, configured; to receive, via the interface, a fundic impedance signal of a subject sensed during an initial calibration period using the at least one implantable fundic sensor of the eating detection device; to receive an indication of actual ingestion events of the subject during the initial calibration period; to perform an analysis of the fundic impedance signal and the actual ingestion events, using a cost function to find each of a plurality of cost function values for each of respective candidate thresholds by; identifying, using an eating detection algorithm, detected ingestion events during the initial calibration period by comparing the fundic impedance signal to the candidate threshold over the initial calibration period, comparing the detected ingestion events to the actual ingestion events to find a number of the detected ingestion events that represent true positives, as number of the detected ingestion events that represent false positives, and a number of the actual ingestion events that represent false negatives, and calculating the cost function value responsively to a measure of sensitivity of the eating detection algorithm and a measure of positive predictive value of the eating detection algorithm, wherein the sensitivity equals the quotient of (a) the number of the true positives divided by (b) is sum of the number of the true positives and the number of the false negatives, and wherein the positive predictive value equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false positives; to set a threshold, by selecting one of the candidate thresholds a the threshold responsively to the cost function values, the crossing of which threshold by the fundic impedance signal is indicative of ingestion; and to store, via the interlace, the threshold in the memory of the eating detection device. - View Dependent Claims (15, 16, 17, 18, 19)
-
-
20. Apparatus comprising:
-
one or more implantable fundic sensors, configured to be applied to a fundus of a subject, and to generate a fundic, impedance signal; and a control element, which comprises a memory, and which is configured to; receive the fundic impedance signal sensed during an initial calibration period, receive an indication of actual ingestion events of the subject during the initial calibration period, perform an analysis of the fundic impedance signal and the actual ingestion events, using a cost function to find each of a plurality of cost function values for each of respective candidate thresholds by; identifying, using an eating detection algorithm, detected ingestion events during the initial calibration period by comparing the fundic impedance signal to the candidate threshold over the initial calibration period, comparing the detected ingestion events to the actual ingestion events to find a number of the detected ingestion events that represent true positives, a number of the detected ingestion events that represent false positives, and a number of the actual ingestion events that represent false negatives, and calculating the cost function value responsively to a measure of sensitivity of the eating detection algorithm and as measure of positive predictive value of the eating detection algorithm, wherein the sensitivity equals the quotient of (a) the number of the true;
positives divided by (b) a sum of the number of the true positives and the number of the false negatives, and wherein the positive predictive value equals the quotient of (a) the number of the true positives divided by (b) a sum of the number of the true positives and the number of the false positives,set a threshold by selecting one of the candidate thresholds as the threshold responsively to the cost function values, the crossing of which threshold by the fundic impedance signal is indicative of ingestion, and store the threshold in the memory. - View Dependent Claims (21, 22, 23, 24, 25, 26)
-
-
27. A computer software product for use with an eating detection device having a memory and at least one implantable fundic sensor, the product comprising a tangible computer-readable medium in which program instructions are stored, which instructions, when read by a computer, cause the computer to receive a fundic impedance signal of a subject sensed during an initial calibration period using the at least one implantable fundic sensor of the eating detection device;
- to receive an indication of actual ingestion events of the subject during the initial calibration period;
to perform an analysis of the fundic impedance signal and the actual ingestion events, using a cost function to find each of a plurality of cost function values for each of respective candidate thresholds by;
(a) identifying, using an eating detection algorithm, detected ingestion events during the initial calibration period by comparing the fundic impedance signal to the candidate threshold over the initial calibration period, (b) comparing the detected ingestion events to the actual ingestion events to find a number of the detected ingestion events that represent true positives, a number of the detected ingestion events that represent false positives, and a number of the actual ingestion events that represent false negatives, and (c) calculating the cost function value responsively to a measure of sensitivity of the eating detection algorithm and a measure of positive predictive value of the eating detection algorithm, wherein the sensitivity equals the quotient of (i) the number of the true positives divided by (ii) a sum of the number of the true positives and the number of the false negatives, and wherein the positive predictive value equals the quotient of (i) the number of the true positives divided by (ii) a sum of the number of the true positives and the number of the false positives;
to set a threshold by selecting one of the candidate thresholds as the threshold responsively to the cost function values, the crossing of which threshold by the fundic impedance signal is indicative of ingestion; and
to store the threshold in the memory of the eating detection device. - View Dependent Claims (28)
- to receive an indication of actual ingestion events of the subject during the initial calibration period;
Specification