Creating and Assessing a Voice Dataset for Automated Classification of Chronic Obstructive Pulmonary Disease

Completed

• Conditions: Chronic Obstructive Pulmonary Disease
• Interventions: Other: COPD

• Registration Number: NCT05897944
• Lead Sponsor: Blekinge Institute of Technology

Brief Summary

This work aims to evaluate whether voice recordings collected from patients diagnosed with COPD and healthy control groups can be used to detect the disease using machine learning techniques.

Detailed Description

Voice data and sociodemographic data on gender and age will be collected through the "VoiceDiganostic" application from the company Voice Diagnostic, which allows one to participate without location dependency. Participants with a diagnosis will be marked as the COPD group, and others will be marked as the healthy control group. Private information such as known comorbidities, personal security numbers, health parameters and communication information will be separately noticed in a participation table for each group.

The collected data will be transformed into mathematical vocal measures called voice features. A dataset consisting of voice features in conjunction with demographics and health data will be constructed for further usage as an input to ML techniques.

Descriptive statistical analysis will be held on attributes containing information on input data and gained outcomes from ML algorithms. The achieved results will be presented in the form of summary tables and graphs.

Study Timeline

• Study Start: 2021-12-16
• Study First Submitted: 2023-06-01
• Study First Submitted QC: 2023-06-01
• Study First Posted: 2023-06-09
• Primary Completion: 2024-08-12
• Study Completion: 2024-10-30
• Status Verified: 2025-03
• Last Update Submitted: 2025-03-17
• Last Update Posted: 2025-03-19

Recruitment & Eligibility

• Status: COMPLETED
• Sex: All
• Target Recruitment: 72

Inclusion Criteria

• being 18 years old and older.

Exclusion Criteria

• being under 18 years old.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
HC	COPD	Participants without Chronic obstructive pulmonary disease diagnosis. Total 38 recruitment, 20 Female, 18 Male
COPD	COPD	Participants with clinically diagnosed Chronic obstructive pulmonary disease. Total 34 recruitment, 18 Female, 16 Male

Primary Outcome Measures

Name	Time	Method
Accuracy	Week 51	Binary detection performance of the ML algorithm
Input data importance scale	Week 51	Features used as input data will be ranked from most important to less important one.

Trial Locations

Locations (1)

Blekinge Institute of Technology; 🇸🇪 Karlskrona, Blekinge, Sweden