Natural Language Processing for Headache Medicine

Headache disorders are diagnosed by clinical history taking and applying the criteria provided within the International Classification of Headache Disorders Third Edition (ICHD-3). To help patients and physicians in making the correct diagnosis, digital technologies based on natural language processing (NLP) approaches may help to identify headache disorders within naturally patient-provided speech. The research aims to develop statistical models through machine-learning NLP applications for the accurate and precise classification of headache disorders with headache expert given ICHD-3 diagnosis as the gold standard. Furthermore, the research also aims to develop statistical models through machine-learning NLP applications for the estimation of impact scores derived from validated headache questionnaires by using texts as input. Patients from the tertiary headache clinic will be recruited to provide oral narrative textual descriptions of their headache attack characteristics and burden of disease related to their headache disorders. The goal of the research is to develop accessible, evidence-based digital medical tools as low-effort applications for the correct diagnosis of headache disorders and estimation of burden of disease due to headache disorders.

Study Overview

• Status: Completed
• Conditions: Migraine; Headache Disorders
• Intervention / Treatment: Other: Natural Language Processing

Detailed Description

Headache disorders are among the most prevalent and disabling conditions worldwide . The Global Burden of Disease study 2016 found migraine to be the second most leading cause of disability worldwide. In the group of 18- to 49-year-olds, migraine is the leading cause of disability . Still, many patients do not receive adequate diagnosis or proper headache-specific treatments.

Physicians performing headache medicine need to have an accurate and complete headache history to construct a correct diagnosis and therapeutic plan. The diagnosis ideally needs to made by applying the International Classification of Headache Disorders Third Edition (ICHD-3). This process is essential to make the correct diagnosis within a reasonable amount of time. However, history taking in headache patients faces many challenges. It heavily relies on oral or written communication between them and patients. It is an effortful and time-consuming practice mostly for non-experienced physicians. Misinterpretation by patients or physicians within dialogue may occur and lead to misunderstandings, wrong diagnosis and maltreatment. Often, patients find difficulties to express all characteristics during a single visit to the doctor, leaving a wealth of useful information for the physician unused. Finally, measuring the burden of disease in headache disorders is difficult and mostly done through validated but rigid questionnaires. It may neglect the often complex but natural impact headache disorders have on all dimensions of human lives.

With the notable exception of e-diaries, digital tools for the headache physician are currently not available. Digital technology may offer many solutions to the challenges stated above. Globally, digitization is expanding faster than before. In the developed world, almost every person now has access to digital tools such as computers, smartphones or tablets. More than 3,5 billion people around the world were estimated to have access to the Internet in 2015 . Artificial intelligence (AI) and machine learning (ML) are entering our digital world rapidly, with already multiple use-cases being implemented in medicine. Algorithms in the field of imaging analysis, speech analysis and electronic patient database mining have been explored already to determine which beneficial effects can be derived from these techniques.

With increased computational speed, storage capacity and evolving user interfaces, new digital clinical applications have potential for helping the patient and physician along the trajectory of dealing with headache disorders. One such field within digital sciences is natural language processing (NLP). It uses text as input to generate mathematical models that have the potential to accurately classify and estimate numeric accounts on the basis of grammar, lexical content, sentimental value of words and word embeddings in sentences.

The investigators believe that the correct application of NLP in headache medicine can ultimately improve lives of many headache sufferers by giving correct diagnosis timely and facilitating communication about the burden of disease between patient and physician. This research project aims to develop NLP tools which are able to analyse patient-produced text about their headache problems to accurately diagnose headache disorders and to estimate the impact of headache disorders on patient's lives.

• Study Type: Interventional
• Enrollment (Actual): 187
• Phase: Not Applicable

Contacts and Locations

Study Locations

• Belgium
  • Ghent, Belgium, 9000
    • University Hospital Ghent
  • Belgie
    • Ghent, Belgie, Belgium, 9000
      • Ghent University Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years and older (Adult, Older Adult)
• Accepts Healthy Volunteers: Yes

Description

Inclusion Criteria:

• Patients visiting the headache clinic of Ghent University Hospital for the first time or in follow up.
• Patients older than 18 years of age.
• Patients should be able to have Dutch as their mother tongue, and be sufficiently able to read, write, understand and speak Dutch.

Exclusion Criteria:

• Patients younger than 18 years of age.
• Patients with a language other than Dutch as mother tongue.
• Patients with substance abuse of alcohol or illicit drugs in the present or past.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Diagnostic
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Participants; Participants of the study	Other: Natural Language Processing; Natural Language Processing: classification and regression tasks.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
F1-scores classification migraine narrative versus cluster headache narrative	Machine learning classification models applying logistic regression, naive bayes classification and support vector machines based on the textual elements of the patients' narratives, to classify the provide narrative as either migraine or cluster headache. Higher F1-scores suggest better classification results.	Baseline

Collaborators and Investigators

• Sponsor: University Hospital, Ghent

Publications and helpful links

General Publications

• Vandenbussche N, Van Hee C, Hoste V, Paemeleire K. Using natural language processing to automatically classify written self-reported narratives by patients with migraine or cluster headache. J Headache Pain. 2022 Sep 30;23(1):129. doi: 10.1186/s10194-022-01490-0.

Study record dates

Study Major Dates

• Study Start (Actual): August 28, 2020
• Primary Completion (Actual): December 31, 2023
• Study Completion (Actual): December 31, 2023

Study Registration Dates

• First Submitted: April 25, 2022
• First Submitted That Met QC Criteria: May 11, 2022
• First Posted (Actual): May 17, 2022

Study Record Updates

• Last Update Posted (Actual): May 16, 2024
• Last Update Submitted That Met QC Criteria: May 14, 2024
• Last Verified: May 1, 2024

More Information

Terms related to this study

• Keywords: Natural Language Processing; Headache Disorders
• Additional Relevant MeSH Terms: Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Pain; Neurologic Manifestations; Headache; Headache Disorders
• Other Study ID Numbers: BC-08263

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No