Wideband Acoustic Immittance

Exploring the WAI Database

Author

Vadym Musiienko

Published

November 24, 2024

Introduction

This analysis explores the Wideband Acoustic Immittance (WAI) Database hosted by Smith College. The database contains measurements of acoustic properties in human ears, including absorbance, power reflectance, impedance, and other related quantities. Implemented in MySQL, this online database serves as a repository for normative adult WAI measures, facilitating data sharing and analysis across different research studies. The database is publicly accessible here.

The analysis has two primary objectives:

  1. Reproduce Figure 1 from Voss (2019), which illustrates mean absorbances across various studies
  2. Visualize WAI mean absorbance across various studies for male and female participats

Import Libraries

library(tidyverse)
library(dbplyr)
library(RMariaDB)

Connect to the database

con_wai <- dbConnect(
  MariaDB(), host = "scidb.smith.edu",
  user = "waiuser", password = "smith_waiDB", 
  dbname = "wai"
)

Data

Firstly, we need to understand the structure of the SQL database, including its tables and the variables within them.

Data base structure

SHOW TABLES;
7 records
Tables_in_wai
Codebook
Measurements
Measurements_pre2020
PI_Info
PI_Info_OLD
Subjects
Subjects_pre2020

The WAI database consists of 7 tables, but our analysis will focus on the three most relevant and up-to-date tables: Measurements, PI_Info, and Subjects.

Measurements table

SELECT *
FROM Measurements
LIMIT 0, 5;
5 records
Identifier SubjectNumber Session Ear Instrument Age AgeCategory EarStatus TPP AreaCanal PressureCanal SweepDirection Frequency Absorbance Zmag Zang
Abur_2014 1 1 Left HearID 20 Adult Normal -5 4.42e-05 0 Ambient 210.938 0.0333379 113780000 -0.233504
Abur_2014 1 1 Left HearID 20 Adult Normal -5 4.42e-05 0 Ambient 234.375 0.0315705 103585000 -0.235778
Abur_2014 1 1 Left HearID 20 Adult Normal -5 4.42e-05 0 Ambient 257.812 0.0405751 92951696 -0.233482
Abur_2014 1 1 Left HearID 20 Adult Normal -5 4.42e-05 0 Ambient 281.250 0.0438399 86058000 -0.233421
Abur_2014 1 1 Left HearID 20 Adult Normal -5 4.42e-05 0 Ambient 304.688 0.0486400 79492800 -0.232931

The Measurements table contains the core data of our analysis, including publication identifiers, instrument information, and most importantly, frequency and absorbance measurements. This table links to PI_Info through the Identifier column and to Subjects through the SubjectNumber column.

PI_Info table

SELECT Identifier, Year, Authors, AuthorsShortList
FROM PI_Info 
LIMIT 0, 5;
5 records
Identifier Year Authors AuthorsShortList
Abur_2014 2014 Defne Abur, Nicholas J. Horton, and Susan E. Voss Abur et al.
Aithal_2013 2013 Sreedevi Aithal, Joseph Kei, Carlie Driscoll, and Asaduzzaman Khan Aithal et al.
Aithal_2014 2014 Sreedevi Aithal, Joseph Kei, and Carlie Driscoll Aithal et al.
Aithal_2014b 2014 Sreedevi Aithal, Joseph Kei, and Carlie Driscoll Aithal et al.
Aithal_2015 2015 Sreedevi Aithal, Joseph Kei, Carlie Driscoll, Asaduzzaman Khan, and Andrew Swanston Aithal et al.

The PI_Info table provides metadata about each publication, including author information and publication year.

Subjects table

SELECT *
FROM Subjects
LIMIT 0, 5;
5 records
Identifier SubjectNumber SessionTotal AgeFirstMeasurement AgeCategoryFirstMeasurement Sex Race Ethnicity LeftEarStatusFirstMeasurement RightEarStatusFirstMeasurement SubjectNotes
Abur_2014 1 7 20 Adult Female Unknown Unknown Normal Normal
Abur_2014 3 8 19 Adult Female Unknown Unknown Normal Normal Session 5 not included do to acoustic leak
Abur_2014 4 7 21 Adult Female Unknown Unknown Normal Normal
Abur_2014 6 8 21 Adult Female Unknown Unknown Normal Normal
Abur_2014 7 5 20 Adult Female Unknown Unknown Normal Normal

The Subjects table contains demographic information about study participants, including ethnicity, sex, and other relevant characteristics.

Part 1: Reproducing Voss (2019) Figure 1

To recreate Figure 1 from Voss (2019), we first need to identify the publications included in the original figure. Let’s examine the available identifiers in the database.

Unique identifiers

SELECT DISTINCT(Identifier) FROM Measurements;
Displaying records 1 - 10
Identifier
Abur_2014
Aithal_2013
Aithal_2014
Aithal_2014b
Aithal_2015
Aithal_2017a
Aithal_2017b
Aithal_2019a
Aithal_2019b
Aithal_2020a

After reviewing the list of Identifiers, I identified the following publications used in Voss (2019): Abur_2014, Feeney_2017, Groon_2015, Lewis_2015, Liu_2008, Rosowski_2012, Shahnaz_2006, Shaver_2013, Sun_2016, Voss_1994, Voss_2010,and Werner_2010.

Data acquisition

To recreate the figure, we need to calculate average absorbance values across frequencies for each publication. The legend requires additional publication information retrieved from the PI_Info table.

Query:

SELECT 
    Frequency,
    AVG(Absorbance) AS Mean_absorbance,
    CONCAT(pi.AuthorsShortList, ' (', pi.Year, ') N=', 
    COUNT(DISTINCT SubjectNumber, Ear), ': ', m.Instrument) AS Publication
FROM Measurements AS m
JOIN PI_Info AS pi ON m.Identifier = pi.Identifier
WHERE m.Identifier IN 
  ("Abur_2014", "Feeney_2017", "Groon_2015", "Lewis_2015", "Liu_2008",
  "Rosowski_2012", "Shahnaz_2006", "Shaver_2013", "Sun_2016",
  "Voss_1994", "Voss_2010", "Werner_2010")
GROUP BY Frequency, m.Identifier, pi.Year, m.Instrument

Custom theme

We designed a custom theme for the plot to align with the style of the original figure.

custom_theme <- theme(
  # Background
  panel.background = element_rect(fill = "white", color = "grey"),
  panel.grid.minor = element_line(color = "gray80"), 
  
  # Axis text and titles
  axis.text = element_text(size = 8),
  axis.title = element_text(size = 11, colour = "black"),
  
  # Legend
  legend.position = c(0, 1),
  legend.justification = c(0, 1),
  legend.key.size = unit(0.4, "cm"),
  legend.text = element_text(size = 9),
  legend.key = element_rect(fill = "white", color = "white"),
  legend.background = element_rect(
    fill = "white",
    color = "grey",
    size = 0.5
  ),
  
  # Title
  plot.title = element_text(hjust = 0.5, size = 14, face = "bold"),
  
  # Ticks on x and y axis
  axis.ticks = element_line(color = "grey", size = 0.5),
  axis.ticks.length = unit(0.2, "cm")
)

Recreate Figure 1 from Voss (2019)

Here is the code for the final visualization:

figure_data |> 
  ggplot(
  aes(Frequency, Mean_absorbance, color = Publication)
  ) +
  geom_line(size = 1) +
  
  # Zoom in on the interval of interest
  coord_cartesian(xlim = c(200, 8000), ylim = c(0, 1)) +
  
  # Add custom ticks on x-axis
  scale_x_continuous(
    breaks = c(200, 400, 600, 800, 1000, 2000, 4000, 6000, 8000),
    trans = "pseudo_log"
                 ) +
  
  # Add custom ticks on y-axis
  scale_y_continuous(
    breaks = c(0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1),
                 ) +
  
  # Add the labels
  labs(
    title = "Mean absorbance from each population in WAI database",
    x = "Frequency (Hz)",
    y = "Mean absorbance",
    color = NULL
  ) + 
  
  # Apply the custom theme
  custom_theme

Voss (2019) describes the figure as follows:

“Mean absorbances for the 12 studies within the WAI database as of July 1, 2019. The legend includes peer-reviewed publications, the number of individual ears, and the equipment used in the studies. For ears with multiple measurements, the average was used. The figure includes all ears in the database, without controlling for the number of ears from each subject.”

Part 2: Mean absorbance for males and females

This analysis aims to determine whether there is a difference in mean absorbance between male and female participants. As in the previous analysis, we use the same set of publications, but this time we examine male and female participants separately.

Data acquisition

Participant sex information is retrieved from the Subjects table, which is joined with the Measurements table to link absorbance data with each subject’s personal information.

Query:

SELECT 
    Frequency,
    CASE 
      WHEN UPPER(Sex) LIKE '%MALE%' THEN 'Male'
      WHEN UPPER(Sex) LIKE '%FEMA%' THEN 'Female'
    END AS Sex,
    AVG(Absorbance) AS Mean_absorbance,
    CONCAT(Sex, ' N=', 
    COUNT(DISTINCT m.SubjectNumber, Ear), ': ', m.Identifier) AS Label
FROM Measurements AS m
JOIN Subjects AS s ON m.SubjectNumber = s.SubjectNumber
WHERE m.Identifier IN ("Abur_2014", "Feeney_2017", "Groon_2015", "Lewis_2015", "Liu_2008", "Rosowski_2012", "Shahnaz_2006", "Shaver_2013", "Sun_2016", "Voss_1994", "Voss_2010", "Werner_2010") 
AND Sex != 'Unknown'
GROUP BY Frequency, s.Sex, m.Identifier

Visualization

The following R code generates the visualization to compare mean absorbance between male and female participants across multiple studies:

# Graph the data
wai_sex |>
  ggplot(
  aes(Frequency, Mean_absorbance, group = Label, color = Sex)
  ) +
  geom_line(size = 1) +
  
  # Zoom in on the interval of interest
  coord_cartesian(xlim = c(200, 8000), ylim = c(0, 1)) +
  
  # Add custom ticks on x-axis
  scale_x_continuous(
    breaks = c(200, 400, 600, 800, 1000, 2000, 4000, 6000, 8000),
    trans = "pseudo_log"
                 ) +
  
  # Add custom ticks on y-axis
  scale_y_continuous(
    breaks = c(0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1),
                 ) +
  
  # Add the labels
  labs(
    title = "Mean absorbance for male and female",
    x = "Frequency (Hz)",
    y = "Mean absorbance",
    color = NULL
  ) + 
  
  # Apply the custom theme
  custom_theme

Each study has two lines: blue representing male participants and red representing female participants. The 12 studies visualized here are the same as those in Part 1.

This plot shows that female participants experience higher absorbance than male participants in the 800 Hz frequency range. However, the difference decreases after approximately 5000 Hz, at which point the absorbance levels for both sexes become nearly identical.

Disconnect from the Database

# Disconnect from the database
dbDisconnect(con_wai, shutdown = TRUE)

Acknowledgments

The normative WAI database and corresponding website are funded by the National Institutes of Health, NIDCD (R15 DC014129). The server is hosted at Smith College.

Citation

Voss, Susan E. Ph.D. Resource Review. Ear and Hearing 40(6):p 1481, November/December 2019. | DOI: 10.1097/AUD.0000000000000790

Wideband Acoustic Immittance (WAI) Database