Presentation is loading. Please wait.

Presentation is loading. Please wait.

An Integrated Computational Model to Diagnose Attention Deficit Hyperactivity Disorder (ADHD) Diane Mitchnick, MSc IS, 2021014 September 27, 2015.

Similar presentations


Presentation on theme: "An Integrated Computational Model to Diagnose Attention Deficit Hyperactivity Disorder (ADHD) Diane Mitchnick, MSc IS, 2021014 September 27, 2015."— Presentation transcript:

1 An Integrated Computational Model to Diagnose Attention Deficit Hyperactivity Disorder (ADHD)
Diane Mitchnick, MSc IS, September 27, 2015

2 Introduction ADHD – Attention Deficit Hyperactivity Disorder (causes inattentiveness, hyperactivity, and impulsiveness) ADHD affects social and learning behavior, response times to certain situations and physical regions of the brain Healthcare analytics used for physical health disorders (brain cancer, diabetes, etc.) [1] Many tools do data mining (collect and filter data through single algorithm) [2] Proposing a combination of common adult ADHD screening tools with different datasets to produce a more effective model Model will use neural networks (running many algorithms in parallel to search for patterns in the data and learn the behavior of classifying the diagnosis) Model can then be used for continuous mental health patient data such as writing data sets (healthcare analytics)

3 Questionnaires / Rating Scales
Current Issues and Questions Issue: Current diagnostic tools for adult ADHD are varied and do not always relate or operate with each other. Research Question: How would an analysis based approach on interacting tools be more beneficial in identifying key areas of an adult ADHD diagnosis than the data-mining based approach on one tool alone? Questionnaires / Rating Scales Adult ADHD Diagnosis Brain Scans CPTs* *Computerized Performance Tests Why I’m looking at doing a meta-analysis on current ADHD screening tools first is because the current tools do not inter-relate or operate with each other. Take for instance a neurological scan from a hospital that has ADHD regions of interest in the brain; it will not coincide with the rating scale administered at the psychologist’s office, because they are two different mediums from two different areas of expertise. Collecting one source of data from one location does not give an effective model for which to base an ADHD diagnosis off of. Which bring us to the research question of this thesis. With the various screening tools for ADHD out there, how would an analysis based approach on interacting tools be more beneficial in identifying key areas of an adult ADHD diagnosis in addition to the data-mining based approach on one tool?

4 Methodology Test Model on Learner Groups, Evaluating Data Questionnaire for tool, using previous verification methods on data Design Software Model Based off of Verification Determine algorithms in for each process based off of data Use Correlation/Causation Probability for Verification Co-variation on method results calculated through ANCOVA tests (in R) Run a Meta-Analysis of ADHD Screening Tools for their Effectiveness Case Study Data from PubMed, EMBASE, and ResearchGate (in Excel) Determine a Constant to Compare the Data Against DSM-V classifier from APA* to determine diagnosis *DSM – Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association)

5 Data to be Collected Behavioural
ASRS (Adult Self Reporting Scale) – 18 question scale Physical Other screening questions to rule out pre-existing conditions (head trauma, learning difficulties, etc.) or comorbidity (anxiety medications that suppress the symptoms of ADHD) CPT (Continuous Performance Testing) - reaction times (possibly Conner CPT) Writing expression – student performance data *Note that all scales have a standard determined for ADHD except written expression

6 Writing expression data to be collected
Formal errors (spelling, grammar) - counted against the norm Morphosyntatic Errors (he/she use, past/present tenses, etc.) - counted against the norm Number of Sentences – counted against the norm Number of words – counted against the norm Connective Cohesion (sentences paired together based on relationship) – percentage complete (ex. intro, body, conclusion) Cohesive Adequacy (incomplete sentences) – percentage complete Time sequence errors (out of sequence in chronological events) – counted against the norm *Based off of Meta-Analysis of 30 papers on ADHD with WLD and writing in students

7 Expected Evaluation and Results
As student writing is continuous and can be collected real-time, the dataset is ideal for testing out the model The model will pre-screen students in any discipline first to determine the presence of adult ADHD, then evaluate the pre-screened student's writing second, making the dataset their writing. A second control group that comes out the first pre-screen (i.e. that do not have a presence of adult ADHD) will be formed and considered the "normal" group. The group's writing will be compared against the first group's writing. Should there be differences, those differences will be analyzed, demonstrating a more effective method of detecting and diagnosing adult ADHD. Patterns in the first group's writing can then be studied for making a new classifier.

8 Targeted Sample Size 500 students
Estimated 400 will pass the first screen of physical questions (no external factors) Estimated 200 will pass the second screen of ASRS Estimated 100 will pass the third screen of CPT reaction times The 100 will be the group that will be tested for writing performance Estimated 75 will have a higher prevalence of ADHD through written expression than the normal group of the 200 students

9 Present and Future Considerations
Pros Application could be used for medical assistance Application could be used as a possible study aid (for nursing, psychology and health studies students analyzing learning behaviour in students) May inspire research on causation and possible prevention of mental disorders Cons Need student's consent to be tested (new research ethics application) May not generate any differences in writing to study Will require a comprehensive literature review (beyond the meta-analysis) and fuzzy logic for analyzing written dataset to determine a standard for the ADHD diagnosis.

10 Questions


Download ppt "An Integrated Computational Model to Diagnose Attention Deficit Hyperactivity Disorder (ADHD) Diane Mitchnick, MSc IS, 2021014 September 27, 2015."

Similar presentations


Ads by Google