Dr. Kashif Sultan

PhD Theme/Topic: Digital Behavioral Biomarkers for ADHD Detection Using Machine Learning and LLM-Based Screen Interaction Analytics

Supervisor: Dr. Kashif Sultan, Associate Professor
Contact #: 0345-5150569
Email: Kashif.buic@bahria.edu.pk
Campus/School/Dept: BUIC H-11/BSEAS/SE
RAC Approved Supervisor for Research Areas: Artificial Intelligence 

Supervisory Record:
PhD Produced: 0
PhD Enrolled: 0
MS/MPhil Produced: 10
MS/MPhil Enrolled: 4

Topic Brief Description:

ADHD is a neurodevelopmental disorder characterized by inattention, impulsivity, and hyperactivity, often affecting academic performance, social behavior, and long-term mental well-being. Traditional diagnostic methods rely on clinical assessments, interviews, and behavioral questionnaires, which can be subjective, time-consuming, and inaccessible for early detection—especially in resource-constrained environments. There is a critical need for an objective, scalable, and technology-driven approach that can support early identification and monitoring of ADHD tendencies.

This research aims to develop an AI-powered prediction system that detects ADHD risk through passive monitoring of screen interaction behavior. By analyzing digital activity patterns such as screen-time duration, app switching frequency, scrolling impulsiveness, and task attention spans, the system will model behavioral markers that align with DSM-5 ADHD symptoms. A Chrome browser extension and mobile application will collect this data and generate a real-time ADHD probability score using machine learning and LLM-based behavioral reasoning. The expected outcome is an accessible, non-invasive early- screening platform that supports parents, educators, and clinicians in recognizing ADHD symptoms earlier and more effectively.

Research Objectives/Deliverables: 

1. Develop an AI-based model for early ADHD detection using passive digital behavioral signals such as screen-time patterns, task switching, scrolling impulsivity, and attention span indicators.
2. Design and deploy a Google Chrome extension and Android application to collect real-time user interaction data and generate an ADHD Risk Score through machine learning and LLM-based analysis.
3. Validate model performance against clinical ADHD assessment tools to ensure accuracy, reliability, and practical applicability for early screening in educational and healthcare settings.

Research Questions:

1. Which screen-interaction behaviors correlate strongly with ADHD diagnostic markers?
2. Can machine learning models classify ADHD vs non-ADHD users using digital passive data?
3. How accurately can an LLM-enhanced platform interpret multimodal user behavior?
4. Can everyday device monitoring reduce the delay in ADHD diagnosis in youth/teens?

Candidate’s Eligibility Profile: 

The candidate must hold an MS/MPhil degree in Computer Science, Software Engineering, AI/Data Science, or a closely related discipline with a minimum CGPA requirement set by the BU for PhD fellowship program. A strong foundation in machine learning, deep learning, and data-driven modeling is essential, with demonstrated skills in Python and relevant AI frameworks. Familiarity with digital behavioral data, mobile/computer interaction analytics, or human-computer interaction research will be considered advantageous.

Experience in research, academic writing, experimental design, and practical implementation of AI models will support successful execution of the project. The candidate should possess strong analytical thinking, problem-solving ability, and the capacity to engage in interdisciplinary work involving psychology, computing, and healthcare. Good communication skills, ethical responsibility in data handling, and interest in assistive/mental health technology are highly desirable.