Research

Five pillars of human-centered AI.

Each pillar carries multiple sub-themes and active projects, connected by one design principle: technical excellence in service of people. Funded projects are catalogued on the Projects & Funding page; full outputs on Publications.

Pillar 01

Health & Healing

Flagship: AI for Health and Healing, a $130,000 TRU HorAIzon Presidential Initiative led by AIRL — including the Wellness Hub, demonstrated live to BC's Minister of State for AI and New Technologies.

Neuro-imaging & brain tumours

A signature line since 2010: the DeepTumor framework, glioma classification with deep features, multiclass texture-statistical models, and the US patent for multi-stage brain-tumour image processing (11,227,387).

Oncology imaging

Breast-cancer detection from whole-slide histopathology, leukemia detection with ensemble CNNs, lung-cancer LDCT screening, ultrasound breast imaging, melanoma segmentation, and an optimization-driven skin-cancer framework (TRU IRF).

Ophthalmology

Diabetic-retinopathy detection from fundus images: MEDCnet memory-efficient CNNs, hybrid deep features, and optimized frameworks developed with student researchers.

Infectious disease

COVID-19 and pneumonia detection from CT and X-ray with optimized deep learning; monkeypox/chickenpox lesion classification; fungal-infection detection with fusion features.

Digital mental health & care education

Privacy-preserving campus LLMs for student support and mental-health navigation; AISimSpeak, an AI simulation platform for nursing education (CAN-Sim); an AI memory-recall assistant for dementia patients.

Secure intelligent health systems

Federated-learning privacy for healthcare IoMT, medical-record security in multimodal images and steganography, blockchain for smart healthcare, tampering detection, and IoT diagnosis platforms (I-CARES).

Pillar 02

Defence & Security

From the governance of AI-generated disinformation to deployed detection systems and a nine-patent document-forensics portfolio.

Disinformation & synthetic media

Detection of AI-written vs. human-written text with deep recurrent networks; sociotechnical research on institutional resilience against synthetic threats in Canada's defence and security sector; misinformation identification for parliamentary information services.

Defence applications

Deep learning for the defence of military installations against UAV-borne attacks (IEEE CSCE); AI decision-support research for defence contexts with international research interns.

Surveillance & threat detection

Weakly-supervised violence detection with transformer attention, smart-city robbery detection, and maritime ship detection from satellite imagery.

Document forensics

Nine granted US patents for writer verification and authorship identification of handwritten text — including damaged and cursive Arabic documents, imaging devices, and digital-pen verification.

Cybersecurity

Malicious URL and PDF detection with machine learning, email fraud analysis, hate/offense speech detection (including low-resource languages), banking cybersecurity, and IoT/WSN security.

Public safety & wildfire resilience

An intelligent wildfire-management system for data collection, prediction and response; a wildfire-risk, air-quality and community-preparedness platform for the Thompson-Okanagan; post-fire reforestation analysis; flood detection from visual imagery.

Pillar 03

Assistive Technologies

Flagships: BlindLine — first among 800 entries at the NEOM OXAGON Hackathon — and the patented deep-learning Braille Reader (US 11,393,361), honoured with the Geneva Gold Medal.

Braille & blind education

"Learning at Your Fingertips" — an IoT-based, AI-powered Braille learning system — plus Braille recognition research and a pan-Canada Braille-learning knowledge-exchange initiative.

Sign-language AI

Deep-learning Arabic Sign Language recognition and translation for automated communication with Deaf and hard-of-hearing individuals, built on the open ArASL dataset (54,049 images).

Accessible interfaces

Eye-gaze user interfaces, real-time IoT voice analysis and smart monitoring for persons with disabilities, and hands-free cloud e-learning environments.

Cognitive support

A vision-and-AI personalized memory-recall assistant for dementia patients, developed through the Mitacs Globalink program.

Inclusive & intelligent education

Generative-AI accessibility adaptation of course content, at-risk student prediction, ML-driven performance assessment, automated grading, retrieval-augmented generation evaluation, and privacy-preserving campus LLMs.

Recognition

Presented to BC Government MLAs on AI in medical diagnostics and assistive technologies; the co-supervised "We'll Care" team won first place and the Audience Choice Award at TRU's Wolves' DEN.

Pillar 04

Indigenous Communities

A growing priority area developed in the spirit of respectful partnership on Tk'emlúps te Secwépemc territory, within Secwepemcúl'ecw.

Language preservation & revitalization

Leveraging AI for the digital preservation and revitalization of Indigenous languages in Canada — building on the lab's depth in low-resource-language NLP and multi-language recognition.

Culturally responsive AI in education

Generative AI for Indigenous student success: culturally responsive custom content creation, presented at TRU's Global Horizons internationalization forum.

Community resilience

Wildfire-risk, air-quality and preparedness platforms serving Interior BC communities, developed with community benefit at the centre.

Roadmap: community-led data sovereignty and stewardship · partnership development with TRU's Office of Indigenous Education · Secwepemctsín-focused language-technology exploration guided by community priorities.

Pillar 05

Applied AI Systems

Cross-cutting intelligent systems for the economy and environment — agriculture, earth science, smart cities, and connected infrastructure.

Agriculture & food

Rice and plant-disease detection with improved CNNs, a global context-aware plant-diagnosis model, cognitive-vision spraying drones, date-fruit classification and disease recognition, fruit-calorie estimation, and marine-life classification.

Earth & environment

Automated mineral-grain segmentation and recognition (FRQNT program with UQAC), the 2026 lithofacies drilling-decision patent (US 12,518,075), seismic-structure classification, oil-spill identification, solar-PV performance prediction, and IoT garbage classification.

Smart cities & mobility

Optimized deep residual networks for Arabic traffic-sign recognition, the autonomous pothole-detection and road-repair patent (US 11,829,959), real-time licence-plate recognition on edge devices, and smart-nursery infant-sound classification.

Connected infrastructure

Secure printer paper-management patent (US 11,748,588), underwater wireless-sensor-network routing, all-optical data-centre interconnects, network traffic engineering, and IoT for oil fields, retail, and smart living.

Document & handwriting AI

Multi-language handwritten numeral and digit recognition, structural-feature methods, scanned-document layout analysis, and automated answer-sheet grading.

Foundations

CNN optimization techniques, GPU-parallel SVMs, memory-efficient architectures, and evaluation methodology — the toolbox powering every pillar.