Method and system for predicting fertility cycle of a user

Abstract

A method (400) and system (100) for predicting a fertility cycle of a user is disclosed A processor (302) acquires physiological biomarker data comprising at least one of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion. A rolling baseline is computed by processing the physiological biomarker data using a rolling-average technique. A stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline. Lifestyle-biomarker data of the user comprising at least one of a body-mass index (BMI), and an age is received. Historical fertility cycle data of the user is retrieved. A fertility cycle comprising an ovulation day and a fertile window of the user is predicted based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical fertility cycle data of the user.

Description

METHOD AND SYSTEM FOR PREDICTING FERTILITY CYCLE OF A USERFIELD OF INVENTION

[0001] The present invention generally relates to the field of health monitoring. More specifically, the present invention relates to a method and system for predicting fertility cycle of a user.BACKGROUND OF THE INVENTION

[0002] The subject matter discussed in the background section should not be assumed to be prior art merely as a result of its mention in the background section. Similarly, a problem mentioned in the background section or associated with the subject matter of the background section should not be assumed to have been previously recognized in the prior art. The subject matter in the background section merely represents different approaches, which in and of themselves may also correspond to implementations of the present technology.

[0003] Accurate prediction of ovulation and the fertile window is an essential aspect of reproductive-health management. Individuals seeking to conceive or manage fertility often rely on traditional approaches such as the calendar method, basal-body-temperature tracking, and ovulation-prediction kits. These conventional techniques, however, depend heavily on manual input and user consistency, and they may fail to accommodate irregular cycles or day-to-day physiological variation.

[0004] Methods based on hormonal test strips can provide indications of ovulation but are limited by chemical stability and handling requirements. Similarly, temperature-based methods only confirm ovulation retrospectively and are influenced by external factors such as illness or ambient temperature. While some modem wearable devices have introduced automatic tracking of physiological signals, these systems frequently rely on community-average data and are unable to adapt to the unique lifestyle, metabolic, and stress patterns of individual users.

[0005] Moreover, menstrual-cycle variability is known to be affected by factors such as age, body-mass index (BMI), and psychological stress. Conventional prediction methods seldom account for the combined influence of these parameters or for longitudinal changes in a user’s physiological baselines. As a result, existing solutions may offer limited precision or reliability in forecasting fertile windows for users with irregular or changing cycles.

[0006] Therefore, there is a need for an improved methodology for predicting fertility cycle of a user.SUMMARY OF THE INVENTION

[0007] This summary is provided to introduce aspects related to a method and system for predicting a fertility cycle of a user, and the aspects are further described below in the detailed description. This summary is not intended to identify essential features of the present subject matter, nor is it intended for use in determining or limiting the scope of the present subject matter.

[0008] A method for predicting a fertility cycle of a user is disclosed. The method may include acquiring, by a processor via one or more sensors of a wearable device worn by the user, physiological biomarker data comprising at least one reading of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion. The method may further include compute, by the processor, a rolling baseline of each of the heart rate, the hear-rate variability, the skin temperature, and the skin perfusion by processing the physiological biomarker data using a rolling-average technique. The method may further include determining, by the processor, a stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline. The method may further include receiving, by the processor, lifestyle-biomarker data of the user comprising at least one of a body-mass index (BMI), and an age. The method may further include retrieving, by the processor, historical cycle data of the user comprising at least one of prior menstrual-cycle start dates, ovulation indicators, or fertile-window records from the wearable device or manually inputted by a user device. The method may further include predicting, by the processor using a prediction model, a fertility time period comprising an ovulation day and a fertile window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical cycle data of the user.

[0009] A system for predicting a fertility cycle of a user is disclosed. The system may include a processor, and a memory communicatively coupled to the processor, the memory storing computer-executable instructions, which when executed by the processor, cause the processor to acquire, via one or more sensors of a wearable device worn by the user, physiological biomarker data comprising at least one of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion. The processor may further compute a rolling baseline byprocessing the physiological biomarker data using a rolling- average technique. The processor may further determine a stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline. The processor may further receive lifestyle-biomarker data of the user comprising at least one of a body-mass index (BMI), and an age. The processor may further retrieve historical cycle data of the user comprising at least one of prior menstrual-cycle start dates, ovulation indicators, or fertile-window records from the wearable device or from a user device. The processor may further predict, using a prediction model, a fertility cycle comprising an ovulation day and a fertile window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical cycle data of the user.

[0010] Other aspects and advantages of the invention will become apparent from the following description, taken in conjunction with the accompanying drawings, illustrating by way of example the principles of the invention.BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings constitute a part of the description and are used to provide a further understanding of the present disclosure.

[0012] FIG. 1 illustrates an exemplary environment of a system for predicting fertility cycle of a user, in accordance with the present disclosure.

[0013] FIG. 2 illustrates an example network architecture of the system of FIG. 1, in accordance with the present disclosure.

[0014] FIG. 3 illustrates a functional block diagram of the wearable device, in accordance with the present disclosure.

[0015] FIGS. 4A and 4B together illustrate a flowchart of a method for predicting a fertility cycle of a user, in accordance with the present disclosure.

[0016] A more complete understanding of the present invention and its embodiments thereof may be acquired by referring to the following description and the accompanying drawings.DETAILED DESCRIPTION OF THE INVENTION

[0017] Exemplary embodiments now will be described with reference to the accompanying drawings. The disclosure may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey its scope to those skilled in the art. The terminology used in the detailed description of the particular exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting. In the drawings, like numbers refer to like elements.

[0018] It is to be noted, however, that the reference numerals used herein illustrate only typical embodiments of the present subject matter, and are therefore, not to be considered for limiting its scope, for the subject matter may admit to other equally effective embodiments.

[0019] The detailed description includes specific details for the purpose of providing a thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without these specific details.

[0020] FIG. 1 illustrates an exemplary environment of a system 100 for predicting fertility cycle of a user, in accordance with the present disclosure. The system 100 may include a wearable device 102, a user device 104, and a user 106 communicatively coupled to each other through one or more wired or wireless connections.

[0021] The wearable device 102 may be configured to be worn by the user 106 on any suitable portion of the body, such as the wrist, arm, or finger. The wearable device 102 may include one or more sensors, for example photoplethysmography (PPG) sensors, temperature sensors, or motion sensors, that continuously or periodically acquire physiological biomarker data of the user 106. Such physiological biomarker data may include, but is not limited to, heart-rate signals, heart-rate variability, skin temperature, and skin perfusion measurements. The wearable device 102 may also include a processor to process and transmit the acquired physiological biomarker data.

[0022] The user device 104 may be implemented as a mobile phone, tablet, or other computing device operable to receive and display information transmitted from the wearable device 102. The user device 104 may execute an application configured to store, analyze, and visualize the user’s physiological and lifestyle data. In some embodiments, the user device 104 maycommunicate with a remote data server or cloud service for additional data processing, prediction modeling, or storage. The user device 104 may further provide notifications or graphical indicators to inform the user 106 of predicted fertility-cycle information or related health insights.

[0023] The user 106 represents an individual whose physiological and lifestyle data are being monitored. The user 106 may interact with the wearable device 102 and the user device 104 by wearing, pairing, or inputting relevant information such as age, body-mass index, or historical cycle dates. The collected information may then be used in subsequent stages of analysis performed by the system 100 to generate personalized fertility predictions.

[0024] FIG. 2 illustrates an example network architecture of the system 100 of FIG. 1, in accordance with the present disclosure. The system 100 may include the wearable device 102, the user device 104, and a data server 204, communicatively coupled through a communication network 202.

[0025] The communication network 202 may include one or more wired or wireless communication links, such as Wi-Fi, Bluetooth, cellular, Radio Frequency (RF) or the Internet, that enable data exchange between the wearable device 102, the user device 104, and the data server 204. The communication network 202 may support transmission of physiological and lifestyle data between system components.

[0026] The user device 104 may operate an application configured to receive, visualize, and synchronize data with the data server 204. The user device 104 may further serve as an intermediary between the wearable device and the data server 204, facilitating local computations or forwarding of encrypted datasets for cloud-based analysis.

[0027] The data server 204 may represent one or more cloud-based or remote computing systems configured to perform large-scale data analytics, maintain historical cycle records, or update prediction models used for fertility forecasting. The data server 204 may store anonymized user information, apply machine -learning algorithms to refine predictive accuracy, and return processed insights to the user device 104 through the communication network 202.

[0028] FIG. 3 illustrates a functional block diagram of the wearable device, in accordance with the present disclosure. The wearable device 102 may be communicatively coupled with thecommunication network 202, the user device 104, and the data server 204 as previously described. The wearable device 102 may include a processor 302, a memory 304, one or more sensors 306 including a photoplethysmography (PPG) sensor 308, an input / output (I / O) device 310, and a network bus 312 that enables communication among these components.

[0029] The processor 302 may include one or more microprocessors, digital signal processors (DSPs), or microcontrollers configured to execute instructions stored in the memory 304. The processor 302 may control overall operation of the wearable device 102, including acquisition of physiological data from the sensors 306, signal conditioning, computation of rollingbaseline values, and wireless transmission of processed data to the user device 104 or the data server 204.

[0030] The memory 304 may include non-transitory computer-readable storage such as flash memory, random access memory (RAM), or other suitable data storage components. The memory 304 may store physiological biomarker data, historical measurements, firmware, and executable instructions for data analysis and communication routines performed by the processor 302.

[0031] The sensors 306 may include various physiological and environmental sensors configured to monitor attributes such as heart rate, skin temperature, motion, or skin perfusion. In an embodiment, the sensors 306 may include the PPG sensor 308, which may emit and detect optical signals to determine volumetric changes in blood flow, thereby enabling computation of heart-rate and heart-rate-variability parameters. The sensors 306 may operate continuously or intermittently under control of the processor 302.

[0032] The I / O device 310 may include a user-interface component such as a touchscreen, button, haptic actuator, or light indicator to receive user inputs or provide feedback regarding system operation. The I / O device 310 may also include connectors or wireless interfaces for pairing the wearable device 102 with the user device 104.

[0033] The network bus 312 may provide a communication pathway interconnecting the processor 302, the memory 304, the sensors 306, and the I / O device 310. The network bus 312 may implement standard data-transfer protocols such as I2C, SPI, or UART to ensure reliable communication among the internal modules of the wearable device 102.

[0034] The wearable device 102 may further communicate with the communication network 202 to transmit collected data to the user device 104 or the data server 204 for storage, further processing, or fertility-cycle prediction.

[0035] By way of an example, the processor 302, via the one or more sensors 306, may acquire physiological biomarker data. The physiological biomarker data may correspond to real-time physiological conditions of the user. The physiological biomarker data may include at least one of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion. In one embodiment, the sensors 306 may include at least one photoplethysmography (PPG) sensor configured to detect volumetric changes in blood flow by emitting light into the skin and measuring the reflected or transmitted light intensity. The PPG sensor may operate at a sampling frequency between approximately 1 Hz and 200 Hz, depending on the measurement mode, to generate a continuous waveform representing pulsatile blood flow. The processor 302 may derive heart rate by detecting successive peaks in the PPG waveform and calculating an average beats-per-minute (BPM) value over a moving time window, for instance, every 10 seconds. In another embodiment, the processor 302 may determine heart-rate variability (HRV) using the time intervals between consecutive heartbeats (R-R intervals) obtained from the PPG waveform. The HRV may be computed through time-domain or frequency-domain techniques, such as the standard deviation of normal-to-normal intervals (SDNN) or the power spectral density of the high-frequency band (0.15-0.4 Hz). The HRV values may provide an indicator of the user’s autonomic nervous system balance, and variations in HRV may be used to assess stress or hormonal phase changes relevant to fertility prediction.

[0036] In some embodiments, a skin-temperature sensor may be included among the sensors 306 to measure peripheral temperature. The sensor may use a thermistor or infrared sensing element to monitor temperature changes within a range of approximately 30°C to 40°C, with a resolution of 0.01 °C. The processor 302 may use these readings to identify subtle thermal shifts associated with changed in cycle and ovulation, typically reflected as a basal temperature increase of 0.2°C to 0.5°C sustained across multiple days. Additionally, the sensors 306 may include an optical or thermal sensor for detecting skin perfusion, defined as the flow of blood through the skin’ s microvascular network. The perfusion level may be determined by analyzing the amplitude and phase variation of the PPG signal or by employing near-infrared (NIR) spectroscopy. The processor 302 may compute a perfusion index by normalizing the ACcomponent of the PPG waveform to its DC baseline, thereby quantifying peripheral circulation variations linked to hormonal cycles or stress responses.

[0037] In an illustrative implementation, the physiological biomarker data may be sampled continuously during both rest and activity periods, and the processor 302 may first filter out unphysiological readings caused by motion artifacts, sensor misalignment, or transient signal loss to enhance the precision of subsequent calculations. .

[0038] The processor 302 may further compute a rolling baseline of each of the heart rate, the heart-rate variability, the skin temperature, and the skin perfusion by processing the physiological biomarker data using a rolling-average technique. The rolling-average technique is a statistical signal-processing technique for processing the physiological biomarker data. In an embodiment, following artifact rejection, the processor 302 may compute rolling-baseline values for each physiological biomarker using a three-day rolling-average window, for example across the preceding day (D-l), the current day (DO), and the following day (D+l). These short-term rolling baselines preserve sensitivity to physiologically meaningful deviations while suppressing noise from day-to-day fluctuations. The resulting baselines may then be used to detect deviations such as elevated resting heart rate, reduced heart-rate variability (HRV), or increased skin temperature, which may correspond to the onset of the ovulatory phase or to stress-induced physiological modulation. In one embodiment, the processor 302 may segment heart-rate measurements obtained from the PPG sensor 308 into fixed-length temporal windows, such as 5-minute or 10-minute intervals, and calculate an average heart rate within each window. The rolling -average technique may involve updating the computed mean each time a new heart-rate sample is acquired, while discarding the oldest sample from the averaging window. This approach allows the baseline to dynamically adapt to gradual physiological changes while filtering transient fluctuations caused by motion or brief stress events.

[0039] In some embodiments, the processor 302 may maintain the rolling baseline over multiple days, for example, 3- to 7-day windows, to establish the user’s resting cardiovascular pattern. The baseline value may be recalculated periodically, such as once per hour or once per day, to capture circadian and hormonal variations associated with different phases of the menstrual or ovulatory cycle. Deviations from this rolling baseline, such as a sustained increase in resting heart rate of more than 3-5 beats per minute, may be interpreted by the processor 302 as indicative of potential ovulation onset or heightened metabolic activity.

[0040] In another embodiment, the rolling-average computation may be applied primarily to suppress statistical outliers and to compensate for minor missing data points in the physiological biomarker dataset. In this implementation, the processor 302 may compute the rolling average over the selected observation window to smooth transient irregularities caused by measurement noise, brief motion artifacts, or incomplete data capture. The averaging ensures that physiologically consistent trends such as gradual heart-rate or temperature variations are preserved while spurious spikes or gaps are neutralized. By maintaining data continuity and stability, this approach enables the processor 302 to generate reliable rollingbaseline values that accurately reflect the user’s underlying physiological state.

[0041] The processor 302 may further determine a stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline. In one embodiment, the processor 302 may continuously monitor heart-rate data derived from the PPG sensor 308 and identify elevations that exceed a predefined deviation threshold, for example, 10% above the computed rolling baseline or a rise of more than 5 beats per minute over a resting baseline maintained for at least 30 seconds. To improve classification accuracy, the processor 302 may distinguish between fitness-related and non-fitness-related elevations by cross-referencing accelerometer and gyroscope readings as well as user-labeled activity events obtained from the sensors 306 or the user device 104. For instance, when the user manually labels an activity such as weight lifting or yoga where wrist motion may be minimal and the gyroscope may not register significant movement the processor 302 may treat these events as fitness -related despite low inertial readings. Conversely, when no corresponding fitness activity is detected through either motion data or user input and the movement intensity remains below a threshold (for example, less than 0.2 g RMS acceleration), the heart-rate elevation may be classified as stress-related rather than exercise-induced. The decay of the stress-rhythm score may be further weighted by the time-of-day associated with each detected non-fitness-related heart-rate elevation to account for diurnal variations in stress sensitivity and recovery.

[0042] In some embodiments, the processor 302 may initialize the stress-rhythm score at a nominal value, such as 100, at the beginning of each 24-hour cycle. Each detected non-fitness- related elevation may trigger a decay function that reduces the score proportionally to the magnitude and duration of the elevation. For example, a mild elevation persisting for less thanone minute may reduce the score by 1-2 points, while a sustained elevation of 10 minutes may reduce it by 5-10 points.

[0043] In another embodiment, the decay of the stress-rhythm score may be weighted by a time-of-day factor associated with each detected heart-rate elevation. The processor 302 may reference an internal circadian model or user-configured diurnal pattern to apply different decay weights for morning, afternoon, and evening intervals. For instance, stress events occurring during late evening hours, such as between 8 PM and midnight, may be weighted more heavily to reflect a greater physiological impact on recovery and hormonal balance. Conversely, events during active daytime hours may result in a reduced decay rate.

[0044] In another embodiment, the processor 302 may periodically restore the stress-rhythm score toward the baseline value in the absence of new stress events, simulating physiological recovery. The recovery rate may be fixed, such as +1 point per hour without stress detections, or adaptive based on prior sleep quality or HRV data stored in the memory 304. The stressrhythm score, or its corresponding rolling -average representation, may then serve as a quantitative feature within the downstream fertility-prediction model, enabling the processor 302 to incorporate both the magnitude and temporal persistence of stress into the probabilistic estimation of the user’s ovulatory and follicular patterns. By integrating this stress-derived feature alongside physiological baselines such as heart-rate variability and skin temperature, the prediction model refines its assessment of cycle timing and variability in response to sustained or transient stress exposures.

[0045] The processor 302 may further receive lifestyle -biomarker data of the user comprising at least one of a body-mass index (BMI), and an age. In one embodiment, the BMI may be computed locally on the user device 104 or on the data server 204 based on user-provided inputs of body weight and height, using the standard relationship BMI = weight (kg) / [height (m)]2, or alternatively, the corresponding weight and height data may be automatically retrieved from loT-enabled measurement tools, such as connected digital scales or body-composition analyzers, integrated with the wearable ecosystem. The computed BMI may then be transmitted to the wearable device 102 through the communication network 202 and stored in the memory 304 for integration with physiological biomarker data.

[0046] In some embodiments, the processor 302 may categorize the BMI value into defined physiological ranges that align with demographic or region- specific health guidelines. Forexample, underweight (BMI < 18.5), normal (18.5 < BMI < 24.9), overweight (25 < BMI < 29.9), and obese (BMI > 30) and associate each category with corresponding ovulatory-cycle patterns derived from population-level datasets. The processor 302 may assign a weighting coefficient to the BMI parameter, reflecting its influence on cycle length variability or ovulation probability as observed in historical data.

[0047] The age of the user may be obtained from user-profile data stored on the user device 104. The processor 302 may use the age value to reference an age-specific fertility profile, such as shorter follicular or luteal phases in users aged 35-40 or extended cycle variability in users under 20 or over 45. In some implementations, both BMI and age may be used as prior inputs to a probabilistic prediction model that adjusts the fertile-window estimation. For example, users with higher BMI values may be assigned a broader fertile window due to observed cycle irregularity, while users within normal BMI and mid-reproductive age may be assigned narrower, higher-confidence windows.

[0048] In another embodiment, the processor 302 may periodically update the lifestylebiomarker dataset to reflect longitudinal changes, such as weight fluctuations or age-related physiological transitions or changes in medication profiles. The processor 302 may receive medication information directly from user input, linked health-record APIs, or integrated electronic health devices. In particular, the system may identify or flag medications known to affect menstrual or ovulatory patterns such as antidepressants, antipsychotic agents, and chemotherapy-related drugs and apply corrective weighting factors to account for their physiological impact. These periodic updates may be synchronized with the data server 204, enabling the fertility-prediction model to continuously recalibrate its parameters and refine predictive accuracy across subsequent cycles.

[0049] The processor 302 may further retrieve historical fertility cycle data of the user from the wearable device 102 or from the user device 104. The historical fertility cycle data may include at least one or prior menstrual-cycle start dates, ovulation indicators, or fertilitywindow records. In one embodiment, the wearable device 102 may store a rolling log of physiological events corresponding to cycle transitions, such as the onset of menstruation, midcycle temperature rises, or characteristic heart-rate fluctuations indicative of ovulation. The processor 302 may retrieve these logs from the memory 304 and compile them into a structural historical dataset for trend analysis. The stored data may be timestamped and indexed by cyclenumber, enabling the processor 302 to compute average cycle length, luteal-phase duration, and variability metrics across multiple cycles.

[0050] In another embodiment, the user device 104 may maintain a synchronized record of user-entered cycle -related events wellness parameters derived from various health-tracking applications integrated through the communication network 202. These may include, but are not limited to, fertility-tracking platforms, nutrition or metabolic -monitoring applications (for example, those tracking ketogenic or intermittent-fasting routines), fitness and activity loggers, or stress and sleep-monitoring tools. The processor 302 may access this data via a secure communication channel through the communication network 202. Retrieved data may include digital entries such as menstrual start and end dates, basal-body-temperature shifts, or positive luteinizing hormone (LH) test results serving as ovulation indicators. The processor 302 may normalize this data into a unified timeline format and align it with the physiological biomarker data captured by the wearable device 102 to enhance prediction accuracy.

[0051] In some embodiments, the processor 302 may analyze historical variability in cycle length and ovulation timing to establish a personalized cycle model for the user. For instance, if historical records show consistent ovulation between days 13 and 15 of a 28-day cycle, the processor 302 may assign a higher probability to fertility predictions occurring within that window in subsequent cycles. Conversely, if prior data indicate high variability (e.g., a standard deviation greater than 3 days in ovulation timing), the processor 302 may broaden the predicted fertile window and lower the associated confidence level.

[0052] The processor 302 may further predict a fertility cycle using a prediction model. The fertility cycle may include an ovulation day and a fertility window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical fertility cycle data of the user. In an embodiment, the prediction model may be executed by the processor 302 may be a machine-learning model (ML) model trained to predict a user’s fertility cycle based on a weighted probabilistic analysis of multimodal inputs, including the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical fertility-cycle data of the user. The ML model may be implemented using one or more algorithms selected from, but not limited to, gradient- boosted decision trees, support-vector machines (SVMs), random-forest ensembles, Bayesian networks, long short-term memory (LSTM) recurrent neural networks, or transformer-based sequence models capable of learning temporal dependencies in biological data streams.

[0053] During a training phase, the model may receive time- stamped datasets collected from a population of users, each dataset including synchronized physiological biomarkers (e.g., heart rate, HRV, skin temperature, perfusion), lifestyle attributes (e.g., BMI, age, stress score), and ground-truth fertility annotations obtained from ovulation test kits or basal-body-temperature logs. The training may employ supervised-leaming techniques with labeled instances of ovulatory and non-ovulatory cycles, enabling the model to learn discriminative patterns between follicular, ovulatory, and luteal phases. Regularization and cross-validation procedures may be applied to prevent overfitting and ensure generalizability across users with different demographics and health baselines.

[0054] In certain embodiments, users may be profiled into demographic personas based on attributes such as age, BMI, and cycle regularity. These personas define population-level priors that inform the model’s baseline assumptions and fallback estimations when user-specific historical lifestyle-biomarker data or historical fertility cycle data are missing. For example, a model trained on aggregated persona data may apply a prior probability distribution of ovulation timing characteristic of users aged 25-30 years with regular 28-day cycles, while applying broader distributions for irregular-cycle personas. Once trained, the prediction model may operate in an inference mode on the user device 104, receiving preprocessed statistical features such as rolling baselines of biomarkers and stress-rhythm indices from the wearable device 102.

[0055] In one embodiment, the prediction model may implement a Bayesian inference framework that assigns posterior probabilities to ovulation events based on multi-modal data inputs. The model may define a likelihood function that correlates the probability of ovulation with time-aligned variations in heart rate, skin temperature, and skin perfusion. The rolling baseline computed earlier may serve as a normalization reference, enabling the detection of subtle follicular-to-luteal transitions. The processor 302 may update the probability distribution dynamically as new sensor data becomes available, thereby refining the predicted fertile window in real time. In some embodiments, additional physiological or environmental data streams such as core body temperature, electrodermal activity, sleep quality, or respiration rate may be acquired through ancillary loT sensors, smart rings, smart scales, or connected health monitors integrated via the communication network 202. The rolling baseline computed earlier may serve as a normalization reference, allowing the detection of subtle follicular-to-luteal phase transitions across these diverse signal modalities. The processor 302 may continuouslyupdate the probability distribution in real time as new sensor data becomes available from one or more connected sources, thereby refining the predicted fertile window with increasing temporal precision and contextual awareness.

[0056] In another embodiment, the prediction model may integrate weighted coefficients that reflect the relative influence of each biomarker type on fertility outcomes. For instance, a stress-rhythm score below a predetermined threshold (e.g., 70 on a 0-100 scale) may be weighted to reduce the probability of ovulation by a fixed percentage, whereas consistent basal temperature elevation may increase the ovulation probability by a similar factor. The weights may be adaptively tuned through machine-learning optimization, where the model retrains itself using the user’s longitudinal data stored in the memory 304 or on the data server 204.

[0057] In some embodiments, the prediction model may further incorporate historical cycle variability parameters, such as average cycle length and standard deviation of ovulation timing, as priors in the probabilistic computation. For example, when prior data indicate a stable 28- day cycle with low variability, the model may assign a narrow fertile-window confidence interval (e.g., days 12-16). Conversely, users with irregular cycles may receive a wider predicted fertile window (e.g., days 10-18) with lower confidence weighting. In certain embodiments, anomalous cycles, including abnormally short or extended cycles resulting from transient stress events, acute illness, or postpartum recovery, may be algorithmically excluded from baseline modeling to preserve the accuracy of probabilistic inference. The system may automatically detect such deviations by applying temporal outlier-detection algorithms, contextual metadata analysis, or cross-correlation with stress-rhythm and lifestyle-biomarker datasets. In an embodiment, the system 100 may also incorporate specialized post-pregnancy return-handling modules, the present embodiment provides a broad, adaptable framework designed to accommodate a variety of physiological and behavioral perturbations without constraining the model architecture to a single method of exclusion or correction.

[0058] In another embodiment, the processor 302 may employ a neural-network-based model trained on population-level fertility datasets to generate initial estimates when individual historical data are insufficient. As the wearable device 102 accumulates multiple cycle records, the prediction model may progressively shift from population-based priors to user-specific parameters, thereby improving predictive accuracy and personalization over time.

[0059] The prediction model may incorporate the user’s stress-rhythm score as a dynamically weighted factor influencing the prediction of ovulation and fertile-window timing. Regular menstrual cyclicity is recognized as a physiological indicator of overall reproductive health, while sustained psychological stress has been correlated with irregular menstruation, anovulatory cycles, and altered luteal-phase duration. The processor 302 may therefore assign a decreasing probability of ovulation when the stress-rhythm score, derived from elevated nonfitness-related heart-rate deviations, remains below a predetermined threshold for a defined temporal window (for example, 48 hours). In some embodiments, transient stress events such as academic, occupational, or environmental stressors may be detected through increased sympathetic activity reflected in elevated resting heart rate or diminished heart-rate variability relative to the computed baseline. Upon detecting such events, the prediction model may simulate suppression of gonadotropin -releasing hormone (GnRH) amplitude, consistent with known stress-induced hypothalamic modulation, thereby widening the predicted fertile window and lowering its confidence weighting. Conversely, recovery of the stress -rhythm score toward its predefined equilibrium value may restore baseline ovulatory probabilities and shorten the high-confidence fertile interval. In further embodiments, the processor 302 may reference population-level datasets demonstrating that women experiencing chronic psychological stress such as pandemic-related, academic, or occupational stress exhibit a statistically higher incidence of menstrual irregularity and abnormal flow patterns. This historical evidence may be encoded into the system’s probabilistic model as an adaptive prior, enabling the algorithm to compensate for stress-induced variability while maintaining predictive robustness. Accordingly, the system 100 may not only predict ovulation and fertility timing but also provide early indicators of potential cycle disruption associated with sustained psychological stress, thereby supporting preventive reproductive-health management.

[0060] The prediction model may further incorporate the user’s body-mass index (BMI) as a modifier influencing the estimated timing and regularity of the fertility cycle. Empirical studies have shown that individuals at both the lower and higher extremes of the BMI spectrum exhibit a higher incidence of irregular, elongated, or anovulatory menstrual cycles when compared to individuals maintaining a normal BMI range. The processor 302 may therefore assign a nonlinear weighting factor to the BMI parameter, such that predicted ovulation probability follows a J-shaped relationship: decreased at both underweight and overweight ranges, and maximized within the normal BMI range (for example, between 18.5 and 22.9 kg / m2).

[0061] In some embodiments, the processor 302 may implement a cubic-spline interpolation model or equivalent non-linear regression function to map BMI to menstrual-cycle variability indices derived from user-specific historical data. For instance, when the BMI value falls below a predefined lower threshold (for example, 18 kg / m2), the model may extend the expected follicular phase length and widen the predicted fertile window to account for delayed ovulation. Conversely, when the BMI exceeds an upper threshold (for example, 25 kg / m2), the system 100 may increase the variance in predicted ovulation timing and reduce confidence weighting to reflect potential anovulatory tendencies commonly associated with elevated adiposity and insulin resistance.

[0062] In another embodiment, the prediction model may employ population-level reference curves established from longitudinal datasets obtained from menstrual-tracking applications. These reference curves demonstrate that individuals with a BMI approximately 20 kg / m2exhibit the most stable cycle length and lowest variability (e.g., -30.5 days average with <1 day standard deviation), whereas both underweight (BMI < 18) and obese (BMI > 30) users show significant deviations. The processor 302 may dynamically compare a user’s real-time BMI calculated from weight and height measurements acquired by the wearable device 102 with these reference relationships to refine probabilistic ovulation estimation. In further embodiments, the system 100 may continuously adjust BMI influence using a personalized adaptive coefficient that updates as the user’s BMI trends toward or away from the optimal range. For example, a sustained reduction of BMI from 27 to 23 kg / m2over multiple logged cycles may cause the prediction model to progressively increase fertility-confidence weighting, reflecting improved hormonal balance and higher ovulation probability. Similarly, a rapid decline of BMI below 17 kg / m2may reduce ovulation confidence and flag potential amenorrhoeic risk.

[0063] Through this adaptive integration of BMI data, the processor 302 not only enhances prediction accuracy of the ovulation day and fertile window but also enables proactive identification of reproductive-health anomalies linked to chronic underweight or obesity. Consequently, the system provides users with actionable insights for maintaining physiological parameters conducive to regular and predictable fertility cycles.

[0064] The prediction model executed by the processor 302 may also incorporate the user’s age as a life- stage covariate that modulates both the expected cycle structure and the uncertainty of the predicted fertility cycle. Age may be represented as a continuous variable (in years) ormapped to age bands (e.g., <20, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, >50). The processor 302 may retrieve the user’s date of birth from the wearable device 102 or the user device 104, compute a current age at the start of each predicted cycle, and select age-conditioned priors for (i) mean cycle length, (ii) cycle-to-cycle variance, (iii) probability of ovulation, and (iv) fertile- window width. For example, priors may encode that cycles in adolescents and very young adults (<20) tend to be longer and more variable (e.g., ~30 days with higher variance), cycles in mid-reproductive ages (e.g., 35-39) tend to be shorter and more regular (e.g., -28-29 days with lower variance), cycles in early perimenopause (e.g., >45) become increasingly irregular, and cycles beyond -50 often lengthen with high variability. These priors act as regularizes when physiological signals are ambiguous, improving robustness of the ovulation-day and fertile-window estimates.

[0065] In some embodiments, the processor 302 employs a piecewise- spline prior over age that jointly parameterizes the follicular-phase length distribution and the ovulation-hazard curve across cycle day. For users <20, the model may broaden the follicular-phase prior and lower the baseline ovulation probability to reflect immaturity of the hypothalamic -pituitary-ovarian (HPO) axis. For users 20-44, the model may narrow the follicular prior and increase ovulation confidence. For users >45, the model may (i) inflate the prior variance of both follicular and luteal phases, (ii) increase the probability mass assigned to anovulatory cycles, and (iii) widen the predicted fertile window, thereby lowering confidence scores unless corroborated by strong physiological biomarkers (e.g., sustained temperature shift, characteristic HRV patterns, or skin-temperature inflection).

[0066] In further embodiments, age interacts with the historical fertility-cycle data and the physiological biomarker data via a Bayesian updating scheme. The processor 302 may initialize an age- conditioned prior for mean cycle length (e.g., po(age)) and cycle variability (e.g., so(age)), and then update these with user-specific posteriors (p*, o*) as additional cycles are logged. For example, a 47-year-old user with five recent cycles showing inter-cycle SD >6 days may cause the model to (a) reduce the prior ovulation probability for the next cycle, (b) expand the fertile-window bounds (e.g., from ±2 days to ±4-6 days around the modal ovulation estimate), and (c) downgrade the fertility-confidence level unless luteal-phase biomarkers indicate a clear biphasic pattern. Conversely, a 33-year-old user with a stable 28-29-day history may trigger tighter priors and higher confidence, provided real-time signals remain consistent with the learned baseline.

[0067] In another embodiment, the processor 302 implements an age-aware ovulation-hazard function h(d I age) that yields the day- specific ovulation probability as a function of cycle day d. The hazard function may be shifted earlier and sharpened for mid-reproductive ages, broadened for <20, and flattened with heavier tails for >45. The predicted fertile window is then obtained by integrating the age-aware hazard with time-aligned biomarker likelihoods (e.g., heart-rate variability minima, skin-temperature rise, and stress-rhythm suppression near the LH surge proxy), producing posterior odds for each day. The fertility-confidence levels described elsewhere (low / high / highest) may be derived from quantiles of these posterior odds; for ages >45, thresholds for “highest” may be elevated to mitigate false positives due to perimenopausal variability. For example, for a user aged 18, the processor 302 may (i) set a longer baseline cycle mean (e.g., ~30 days), (ii) apply a larger variance prior, and (iii) require stronger corroboration from rolling-baseline deviations and temperature shifts before assigning a “highest” confidence period. For a user aged 38, the processor 302 may (i) assume -28-29 day cycles with lower variance, (ii) position the ovulation -hazard peak narrowly around cycle days 13-15, and (iii) produce a compact fertile window with elevated confidence. For a user aged 51, the processor 302 may (i) extend the prior cycle length (e.g., -3O-31 days) with high variance, (ii) increase the prior probability of anovulation, and (iii) expand the fertile window but cap confidence unless biomarker evidence strongly supports a luteal shift.

[0068] In still further embodiments, age is combined with BMI and stress-rhythm score as multiplicative or additive modifiers to the age-conditioned priors, enabling the processor 302 to model compounded effects e.g., a 46-year-old with elevated BMI and high stress receives widened fertile-window bounds and reduced ovulation probability unless individualized data contradicts the compounded risk. This integrated, age-aware design improves alignment of the predicted ovulation day and fertile window with physiologic realities across the reproductive lifespan while maintaining calibrated confidence outputs for user notification on the wearable device 102 and / or the user device 104.

[0069] The processor 302 may further categorize the ovulation day within the fertile window into fertility-confidence levels that may include at least a low-confidence period, a high- confidence period, and a highest-confidence period.

[0070] In one embodiment, the processor 302 may implement a confidence-level assignment module that derives fertility-confidence levels directly from the probability outputs of the trained prediction model. Rather than relying on fixed, rule -based thresholding, the modelintrinsically learns to associate each day within the fertile window with a corresponding likelihood of ovulation occurrence based on multi-dimensional feature correlations observed during training. The prediction model, such as a Bayesian neural network, probabilistic gradient-boosted ensemble, or Gaussian process regressor, may output posterior probability scores representing the estimated confidence of ovulation for each day in the cycle.

[0071] The confidence levels including low, high, and highest confidence tiers may thus emerge naturally from the model’s probability distributions rather than being predefined. For example, the model may assign a relatively low probability to early follicular-phase days, an increasing probability to peri-ovulatory days, and a declining probability during the luteal phase. The processor 302 may interpret these probabilistic gradients to categorize each day within the fertile window according to its inferred likelihood, thereby creating dynamic confidence tiers that evolve with continuous sensor inputs and contextual updates.

[0072] In certain embodiments, the model’s probability calibration may be refined using historical cycle data and sensor consistency metrics, such that users exhibiting stable physiological patterns receive narrower, high-confidence windows, whereas users with irregular biomarkers or intermittent data receive broader, lower-confidence ranges. This enables adaptive personalization, ensuring that the fertility-confidence levels reflect both current model inference and cumulative behavioral consistency. During training, the model learns these probability boundaries implicitly through exposure to large-scale labeled fertility datasets, eliminating the need for arbitrary rule-based cutoffs.

[0073] In another embodiment, the categorization may incorporate multi-feature scoring, where individual physiological biomarkers contribute separate confidence components. For instance, consistent skin-temperature elevation, reduced heart-rate variability, and stable stressrhythm scores over a 24-hour interval may together yield a composite high-confidence score for ovulation. Conversely, the presence of irregular heart-rate spikes or stress-related deviations from baseline may reduce the daily confidence level for fertility prediction.

[0074] The processor 302 may further notify the user of the predicted fertility cycle and the corresponding fertility-confidence levels via at least one of the wearable device 102 or the user device 104. In one embodiment, the wearable device 102 may include a display interface 308 configured to present the predicted fertile-window dates and ovulation day in a graphical or textual format. The processor 302 may generate a visual timeline or calendar layout in whicheach day within the fertile window is color-coded according to its assigned confidence level for example, yellow for a low-confidence period, orange for a high-confidence period, and red for the highest-confidence period. In some embodiments, the display interface 308 may further include icons, vibration patterns, or haptic cues to discreetly notify the user of approaching fertile phases or expected ovulation. In another embodiment, the user device 104 which may be a smartphone, tablet, or computing terminal paired with the wearable device 102 may execute a companion application configured to receive and render fertility predictions transmitted via the communication interface 310. The processor 302 may encrypt and transmit the prediction data, along with corresponding confidence values and timestamps, through a secure wireless link such as Bluetooth Low Energy (BLE), Wi-Fi, or LTE. Upon receipt, the user device 104 may present the data in the form of notifications, dashboard widgets, or interactive analytics panels, allowing the user to view fertility trends, stress patterns, and prior cycle summaries.

[0075] FIGS. 4A and 4B together illustrate a flowchart 400 of a method for predicting a fertility cycle of a user, in accordance with the present disclosure. The method may be executed by the processor 302 of the wearable device 102. The method may include a plurality of steps, so as to predict a fertility cycle of a user.

[0076] At step 402, physiological biomarker data may be acquired via one or more sensors of a wearable device worn by the user. The biomarker data may include at least one of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion.

[0077] At step 404, a rolling baseline may be computed by processing the physiological biomarker data using a rolling-average technique.

[0078] At step 406, a stress-rhythm score that decays from a predefined value may be determined in response to non-fitness-related heart-rate elevations detected as deviations from the heart-rate rolling technique. The decay of the stress-rhythm score may be weighted by a time-of-day associated with each detected non-fitness-related heart-rate elevation.

[0079] At step 408, lifestyle -biomarker data of the user may be received. The lifestylebiomarker data may include at least one of a body-mass index (BMI), and an age.

[0080] At step 410, historical fertility cycle data of the user may be retrieved from the wearable device or from a user device. The historical fertility cycle data may include at least one of prior menstrual-cycle start dates, ovulation indicators, or fertile-window records.

[0081] At step 412, a fertility cycle may be predicted using a prediction model. The fertility cycle may include an ovulation day and a fertile window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestylebiomarker data, and the historical fertility cycle data of the user.

[0082] At step 414, the ovulation day within the fertile window may be categorized into fertility-confidence levels that may include at least a low-confidence period, a high-confidence period, and a highest-confidence period.

[0083] At step 416, the user may be notified of the predicted fertility cycle and the corresponding fertility-confidence levels via at least one of the wearable device or the user device.

[0084] Thus, the method 400 and the system 100 overcome the limitations of conventional fertility-tracking techniques that rely solely on calendar calculations, hormone test kits, or retrospective basal-body-temperature confirmation. By integrating real-time physiological biomarkers, such as heart rate, heart-rate variability, skin temperature, and skin perfusion, with lifestyle indicators including stress-rhythm score, body-mass index, and age, the disclosed system 100 enables a personalized, adaptive, and continuously updated prediction of the user’s ovulation day and fertile window. Furthermore, by employing rolling-baseline analysis, probabilistic modelling, and dynamic confidence classification, the system 100 ensures higher precision and reliability even for users with irregular cycle or incomplete historical data.

[0085] In addition, the disclosed architecture leverages machine-executable models implemented on the wearable device 102 and the user device 104, thereby enabling on-device analytics, reduced latency, and improved data privacy. The capability to integrate both individual-specific and population-level fertility data allows the system to adaptively calibrate its predictions over time, providing a robust, non-invasive, and user-centric approach to fertility-cycle forecasting. Accordingly, the embodiments described herein offer a significant advancement over traditional ovulation prediction systems by enabling proactive reproductive- health monitoring and predictive decision support within a secure, connected wearable ecosystem.

[0086] This does not necessarily imply that each such reference is to the same embodiment(s), or that the feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embodiments.

[0087] As used herein, the terms “include”, “comprises”, “including” and / or “comprising” when used in this specification, specify the presence of stated features, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof. It will be understood that when an element is referred to as being “connected” or “coupled” to another element, it can be directly connected or coupled to the other element or intervening elements may be present. Furthermore, “connected” or “coupled” as used herein may include operatively connected or coupled. As used herein, the term “and / or” includes any and all combinations and arrangements of one or more of the associated listed items.

[0088] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure pertains. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

Claims

We claim:

1. A method (400) for predicting a fertility cycle of a user, comprising: acquiring (402), by a processor (302) via one or more sensors (306) of a wearable device (102) worn by the user, physiological biomarker data comprising at least one reading of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion; computing (404), by the processor (302), a rolling baseline of each of the heart rate, the heart-rate variability, the skin temperature, and the skin perfusion by processing the physiological biomarker data using a rolling-average technique; determining (406), by the processor (302), a stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline; receiving (408), by the processor (302), lifestyle -biomarker data of the user comprising at least one of a body-mass index (BMI), and an age; retrieving (410), by the processor (302), historical fertility cycle data of the user from the wearable device (102) or from a user device (104), the historical fertility cycle data comprising at least one of prior menstrual-cycle start dates, ovulation indicators, or fertile- window records; and predicting (412), by the processor (302) using a prediction model, a fertility cycle comprising an ovulation day and a fertile window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical fertility cycle data of the user.

2. The method (400) as claimed in claim 1, wherein the decay of the stress-rhythm score is weighted by a time-of-day associated with each detected non-fitness-related heart-rate elevation.

3. The method (400) as claimed in claim 1, comprising:categorizing (414), by the processor (302), the ovulation day within the fertile window into fertility-confidence levels comprising at least a low-confidence period, a high-confidence period, and a highest-confidence period.

4. The method (400) as claimed in claim 1, comprising: notifying (416), by the processor (302), the user of the predicted fertility cycle and the corresponding fertility-confidence levels via at least one of the wearable device (102) or the user device (104).

5. A system (100) for predicting a fertility cycle of a user, the system (100) comprising: a processor (302); and a memory (304) communicatively coupled to the processor (302), the memory (304) storing computer-executable instructions, which when executed by the processor (302), cause the processor (302) to: acquire, via one or more sensors (306) of a wearable device (102) worn by the user, physiological biomarker data comprising at least one of a heart rate, a heart-rate variability, a skin temperature, and a skin perfusion; compute a rolling baseline rolling baseline of each of the heart rate, the heartrate variability, the skin temperature, and the skin perfusion by processing the physiological biomarker data using a rolling -average technique; determine a stress-rhythm score that decays from a predefined value in response to non-fitness-related heart-rate elevations detected as deviations from the rolling baseline; receive lifestyle -biomarker data of the user comprising at least one of a bodymass index (BMI), and an age; retrieve historical fertility cycle data of the user from the wearable device (102) or from a user device (104), the historical fertility cycle data comprising at least one of prior menstrual-cycle start dates, ovulation indicators, or fertile-window records; andpredict, using a prediction model, a fertility cycle comprising an ovulation day and a fertile window of the user based on a weighted probabilistic analysis of the physiological biomarker data, the stress-rhythm score, the lifestyle-biomarker data, and the historical fertility cycle data of the user.

6. The system (100) as claimed in claim 1, wherein the decay of the stress-rhythm score is weighted by a time-of-day associated with each detected non-fitness-related heart-rate elevation.

7. The system (100) as claimed in claim 1, wherein the processor-executable instructions, when executed by the processor (302), cause the processor (302) to: categorize the ovulation day within the fertile window into fertility-confidence levels comprising at least a low-confidence period, a high-confidence period, and a highest- confidence period.

8. The system (100) as claimed in claim 1, wherein the processor-executable instructions, when executed by the processor (302), cause the processor (302) to: notify the user of the predicted fertility cycle and the corresponding fertility-confidence levels via at least one of the wearable device (102) or the user device (104).

Images