Getting to Know A/Prof Catherine Shi

Tell us about yourself

Kia ora koutou. I’m Catherine Shi, AUT’s contact for the Healthcare AI Research Network and Head of the new Department of Data Science & Artificial Intelligence at AUT. My work sits at the intersection of computer science and healthcare, helping teams turn algorithms into tools that make a difference for patients and clinicians. I’ve been involved in projects such as brain MRI quantification, LLMs for epilepsy care, and an AI-enabled stroke- prevention pathway.

What are your areas of interest?

I am interested in innovative software solutions in healthcare AI. In particular, my work centres on computational methods for interpreting neurological images. My recent interests extend to multimodal models (imaging, text, and other clinical parameters), foundation models/LLMs for clinical documentation and decision support. I’m also keen on low-cost, point-of-care screening that can be used in primary care and rural settings.

Tell us about projects in your new department?

At AUT’s Department of Data Science & AI, our healthcare-related research work spans sensing, decision support and deployment. A/Prof. Sam Madanian leads digital-health projects that redesign care pathways with data science, telehealth and decision support, bridging research and front-line practice. A/Prof. Sira Yongchareon develops AI for pervasive and IoT environments with human-centred sensing, activity recognition and edge intelligence that enable remote monitoring and ambient assisted living. Prof. William Wong is a pioneer in cognitive engineering and visual analytics, designing interactions that improve clinical situation awareness and safety-critical decision-making, which provide foundations for usable, trustworthy AI tools. A/Prof. Boris Bacic applies computer vision and computational intelligence to human-motion analysis for rehab and health, building explainable, privacy-aware feedback systems relevant to gait and fall-risk monitoring. Dr. Yanbin Liu works on robust deep learning under data scarcity and domain shift (few-shot/meta-learning), which is key to generalisable medical AI across sites and populations. Dr. Anuradha Singh brings embedded systems, wireless sensing and biosignal processing expertise to low-power wearables and continuous health monitoring at the edge.

What opportunities do you see for AI to have a real impact in health?

LLM “copilots” for timely, reliable clinical advice.

Well-designed large language models can

sit inside the clinical workflow, which summarise notes, surface red flags, and suggest guideline-aligned differentials and next steps with citations. The value is speed plus consistency: faster access to best-practice advice, clearer rationales, and calibrated uncertainty so clinicians can judge when to trust, verify, or escalate. Guardrails matter: local pathways, audit trails, and bias/quality monitoring baked in.

Precision medication from real-world data.

By learning from large, linked datasets (registries, EHRs, imaging, wearables), AI can estimate treatment effects for “patients like mine,” improving choices of drug, dose, and sequencing. Think epilepsy, cardiometabolic disease, or polypharmacy in older adults—where personalised risk–benefit matters. The opportunity is a learning health system: model → recommendation → outcome feedback → better model, evaluated with external validation and pragmatic trials.

Transferring specialist knowledge to primary care for prevention and rehab.

AI can bring specialist-grade support to GP clinics and community settings—smarter triage and screening (e.g., portable imaging, simple sensors), culturally safe self-management tools, and personalised rehab plans that adapt to progress at home. This lowers access barriers, shortens time-to-intervention, and helps address inequities across Aotearoa by putting high-quality guidance where people actually receive care.

Which emerging AI technologies or trends are you most excited about?

I’m passionate about the following trends in healthcare AI in Aotearoa.

• Multimodal foundation models that natively combine imaging, text and signals (retinal photos, carotid ultrasound, EEG/wearables) to give one calibrated answer with explanations clinicians can trust.

• On-device/edge AI running on portable scanners and phones, so screening and rehab support can happen in primary care and rural clinics with low latency and stronger privacy.

• Causal and counterfactual analytics on real-world data to move from “best average drug” to “best drug for this person,” including dose/sequence recommendations and uncertainty estimates.

• Federated learning and governed synthetic data, letting hospitals collaborate without moving data, which is critical for Māori and Pacific data sovereignty and for robust external validation.

For collaborations or queries, Catherine can be contacted on catherine.shi@aut.ac.nz