Techniques for predicting menstrual cycle onset

A method for detecting a menstrual cycle onset, comprising:

• causing a graphical user interface of a user device to display a request to activate a menstrual cycle tracking mode of a health-related application;
• receiving, in response to the request, a first user input indicating activation of the menstrual cycle tracking mode;
• receiving, from a wearable device configured to be worn on a finger of the user, physiological data measured from a user by the wearable device, the physiological data comprising at least distal temperature data and heart rate data, wherein the wearable device comprises one or more temperature sensors configured to interface with a skin surface of the finger of the user to measure the distal temperature data continuously for at least a plurality of days, and wherein the distal temperature data captures a higher degree of body temperature fluctuations as compared to a core temperature associated with the user;
• training a machine learning classifier in a first training stage to estimate a first future menstrual cycle onset day that the user experiences a first day of a menstrual cycle based at least in part on inputting the physiological data into the machine learning classifier and inputting one or more user inputs confirming a previous menstrual cycle onset day;
• filtering out at least a first portion of the received distal temperature data corresponding to a time interval for which the received heart rate satisfies a threshold;
• fitting a second portion of the received distal temperature data measured for at least the plurality of days to a trigonometric or polynomial function comprising a plurality of features based at least in part on filtering out at least the first portion of the received distal temperature data, wherein the second portion of the received distal temperature data excludes the first portion of the received distal temperature data based at least in part on the filtering;
• receiving, via the user device, an indication of a corresponding plurality of menstrual cycle onset days for at least a plurality of months after fitting the second portion of the received distal temperature data to the trigonometric or polynomial function and an indication of a respective set of one or more symptoms experienced by the user prior to at least a subset of the plurality of menstrual cycle onset days;
• calculating, after receiving the indication of the corresponding plurality of menstrual cycle onset days, a duration between the plurality of features and the corresponding plurality of menstrual cycle onset days based at least in part on fitting the second portion of the received distal temperature data to the trigonometric or polynomial function;
• identifying one or more trends between the sets of one or more symptoms and the at least subset of the plurality of menstrual cycle onset days;
• estimating, using the machine learning classifier, a range of first future menstrual cycle onset days based at least in part on applying the duration to a most recent feature of the trigonometric or polynomial function;
• causing a graphical user interface of the user device to display a user interface page of a health-related application, the user interface page comprising an indication of the range of first future menstrual cycle onset days, an indication of a current day of a current menstrual cycle of the user, an indication of a current menstrual cycle phase of the current menstrual cycle, and a graphical representation of the second portion of the received distal temperature data that illustrates a change associated with a first temperature value of the user associated with a previous day of the plurality of days and a second temperature value of the user associated with a current day;
• causing the graphical user interface to display one or more activity recommendations for the user, wherein the one or more activity recommendations are based at least in part on the current menstrual cycle phase and on the physiological data;
• receiving, via the graphical user interface of the user device, a first user input comprising an indication of one or more current symptoms experienced by the user;
• estimating, using the machine learning classifier, the first future menstrual cycle onset day within the range of first future menstrual cycle onset days based at least in part on the one or more current symptoms and the one or more trends;
• receiving, via the graphical user interface of the user device, a second user input comprising an indication of an actual menstrual cycle onset day that confirms or denies whether the user experienced the first future menstrual cycle onset day;
• retraining the machine learning classifier in a second training stage to estimate a plurality of future menstrual cycle onset days based at least in part on inputting both the first user input comprising an indication of one or more current symptoms experienced by the user and the second user input comprising the indication of the actual menstrual cycle onset day into the machine learning classifier;
• and
• estimating, using the machine learning classifier, the plurality of future menstrual cycle onset days based at least in part on retraining the machine learning classifier in the second training stage.