Characterization and analysis of self-tracked menstrual cycle data

Work on the characterization and analysis of menstrual cycles using self-tracked mobile health data

We provide a conda environment file for ease of replication in ./menstrual_cycle_analysis.yml

Present directories

doc

• doc/characterization: manuscripts on "Characterizing physiological and symptomatic variation in menstrual cycles using self-tracked mobile-health data"

  • Characterizing physiological and symptomatic variation in menstrual cycles using self-tracked mobile health data. Li, K.; Urteaga, I.; Shea, A.; Vitzthum, V.; Wiggins, C. H; and Elhadad, N. Nature Partner Journal Digital Medicine, 3(79), 2020.

• doc/prediction: manuscripts on predictive models:

  • A generative, predictive model for menstrual cycle lengths that accounts for potential self-tracking artifacts in mobile health data. Li, K.; Urteaga, I.; Shea, A.; Vitzthum, V.; Wiggins, C. H; and Elhadad, N. In NeurIPS 2020 Workshop Machine Learning for Mobile Health, 2020. Contributed Talk.
  • A generative, predictive model for menstrual cycle lengths that accounts for potential self-tracking artifacts in mobile health data. Li, K.; Urteaga, I.; Shea, A.; Vitzthum, V. J.; Wiggins, C. H.; and Elhadad, N. arXiv e-print:2102.12439.
  • A predictive model for next cycle start date that accounts for adherence in menstrual self-tracking. Li K.; Urteaga, I.; Shea, A.; Vitzthum, V. J.; Wiggins, C. H.; and Elhadad, N.; Journal of the American Medical Informatics Association, Volume 29, Issue 1, Pages 3–11, January 2022.
  • A Generative Modeling Approach to Calibrated Predictions:A Use Case on Menstrual Cycle Length Prediction. Urteaga, I.; Li, K.; Wiggins, C.; and Elhadad, N. In Proceedings of the 6th Machine Learning for Healthcare, 2021.

src

Main directory with source code utilities.

• src/characterization
  Directory with code for data processing

• src/prediction
  Directory with code for predictive modeling and evaluation.

scripts

Main directory with scripts to run, evaluate and plot experiments.

Expected directory structure and content

data

Cycle length only information for predictive work

• ./data/cycle_length_data/cycle_lengths.npz
  Numpy array with I (number of individuals) by C (number of cycles per-individual) information

preprocessed_data

Pre-processed dataframes with cycles and tracking data were used for the characterization of menstrual cycles using self-tracked mobile health data: these are not publicly available.

results

Directory for plots and results

Characterization outputs for code in src/characterization

• ./results/characterizing_cycle_and_symptoms
  Results regarding the initial exploratory analysis to characterize the menstrual cycle and self-tracked symptoms

• ./results/characterizing_cycle_and_symptoms/cohort_summary_statistics
  Summary statistics and plots for the npjDigitalMedicine cohort

• ./results/characterizing_cycle_and_symptoms/cycle_period_length_analysis
  Summary statistics and plots regarding the npjDigitalMedicine cohort's self-reported cycles

• ./results/characterizing_cycle_and_symptoms/symptom_tracking_analysis_bootstrapping_{nbootstrapped}
  Results for a bootstrapped analysis (with nbootstrapped samples) of the npjDigitalMedicine cohort's self-tracked symptoms

Predictive outputs

• ./results/evaluate_predictive_models/
  Directory for results per each evaluated cycle length dataset and model