README.md

ThereWillBeBeeps

---

Predictable sequential structure augments auditory sensitivity at threshold

Nadège Marin¹, Grégory Gérenton¹, Hadrien Jean², Nihaad Paraouty², Nicolas Wallaert², Diane S. Lazard^1,3, Keith B. Doelling^1*, Luc H. Arnal^1*

1. Université Paris Cité, Institut Pasteur, AP-HP, Inserm, Fondation Pour l'Audition, Institut de l’Audition, IHU reConnect, F-75012 Paris, France
2. My Medical Assistant SAS, Reims, F-51100, France
3. Princess Grace Hospital, ENT & Maxillo-facial Surgery Department, Monaco

^* These authors jointly supervised this work

Abstract

Human hearing is highly sensitive and allows us to detect acoustic events at low levels. However, sensitivity is not only a function of the integrity of cochlear transduction mechanisms, but also constrained by central processes such as attention and expectation. While the effects of distraction and attentional orienting are generally acknowledged, the extent to which probabilistic expectations influence sensitivity at threshold is not clear. Classical audiometric tests, commonly used to assess hearing thresholds, do not distinguish between bottom-up sensitivity and top-down processes. In this study, we aim to decipher the influence of various types of expectations on hearing thresholds and how this information can be used to improve the assessment of hearing sensitivity. Our results raise important questions regarding the conventional assessment of hearing thresholds, both in fundamental research and in audiological clinical assessment.

---

This repo

The project is organized as Jupyter notebooks, which contain the steps and output of the analysis. These notebooks call functions located in separate Python files.

Usage

Notebooks are meant to be run in increasing order (1_, 2_, etc), starting with preprocessing and then analysis. Each notebook contains detailed comments about what each section of code is doing.

The preprocessing and analysis directories contain subdirectories named funcs, which contain Python scripts for various functions that the notebooks call.

Note: you need an extra python script API_access.py with private access codes for the automated pure-tone audiometry API to run some of the code (preprocessing 2-3, analysis 2).

The free-est way to run these notebooks is with jupyter notebook. Here is one way to load this repo to run smoothly

Clone code and move into folder

git clone git@gitlab.pasteur.fr:ida-public/PredAudio.git
cd PredAudio

Create conda environment and install necessary packages

conda create -n beepbeep python=3.10
conda activate beepbeep
pip install -r requirements.txt
conda install jupyter

Include project directory as path in conda env and deactivate environment to go do something else.

conda develop .
conda deactivate

Then when you are ready to run your notebooks just do the following while in the project directory:

conda activate beepbeep
jupyter notebook

Requirements

This repo relies on commonly used python packages for data analysis and visualization, as well as the 'requests' module for communication with the pure-tone audiometry API.

See requirements.txt for the detailed short list.

---

Pipeline

Preprocessing

• Compute audiograms for the 3-AFC task ── 1_3AFC_thresholds
• Fix responses in the Continuous task ── 2_fix_continuous_responses
• Resample audiograms ── 3_resample_audiograms

Analysis

• Compare average thresholds ── 1_threshold_analysis
• Compute the global random audiogram ── 2_global_random_audiogram
• Plot example data to visually explain p50 method ── 3_fig2C_p50_example_data
• Compare average p50 values ── 4_p50_analysis
• Check performance correlation across conditions ── 5_performance_correlation
• Cluster paradigms based on performance ── 6_clustering_paradigms
• Compute false alarm rates ── 7_catch_trials
• Correlate performance with false alarm rate, musical sophistication & age ── 8_linear_models

---

Repository structure:

.
├── data
│   ├── audiograms
│   ├── dataframes
│   └── raw_data
│       └── one folder/participant
│           ├── Bayesian
│           ├── Continuous
│           ├── Cluster
│           └── 3AFC
│
├── figures
│
├── README.md
├── requirements.txt
│
└── scripts
    ├── __init__.py
    ├── figure_params.py
    ├── stats.py
    ├── utils.py
    ├── API_access.py **** PRIVATE ****
    │
    ├── preprocessing
    │   ├── __init__.py
    │   │
    │   ├── funcs
    │   │   ├── __init__.py
    │   │   ├── fix_continuous_responses.py
    │   │   └── resample_audiograms.py
    │   │
    │   ├── 1_3AFC_thresholds.ipynb
    │   ├── 2_fix_continuous_responses.ipynb
    │   └── 3_resample_audiograms.ipynb
    │
    └── analysis
        ├── __init__.py
        │
        ├── funcs
        │   ├── __init__.py
        │   ├── global_random_audiogram.py
        │   ├── linear_models.py
        │   ├── p50_analysis.py
        │   ├── performance_correlation.py
        │   ├── plots.py
        │   └── threshold_analysis.py
        │
        ├── 1_threshold_analysis.ipynb
        ├── 2_global_random_audiogram.ipynb
        ├── 3_fig2C_p50_example_data.ipynb
        ├── 4_p50_analysis.ipynb
        ├── 5_performance_correlation.ipynb
        ├── 6_clustering_paradigms.ipynb
        ├── 7_catch_trials.ipynb
        ├── 8_linear_models.ipynb
        └── SupplFig_example_data.ipynb  - in construction