Rare and neurodevelopmental conditions affect 5–8% of the population, yet children with these conditions routinely wait years for a confirmed diagnosis. The problem is structural: diagnostic data (clinic notes, imaging scans, speech recordings, behavioural assessments, genetic phenotypes) exist in separate systems, assessed by different specialists at different times, with no way to integrate the signals into a coherent picture.
Artificial intelligence offers a credible path forward, but most current approaches analyse one data type at a time and are tested only in controlled research settings. What is needed is a framework that works across data modalities simultaneously, explains its reasoning in terms clinicians can act on, and holds up across different healthcare systems and patient populations.
This project develops and validates a multimodal AI framework for earlier detection and stratification of neurodevelopmental and rare conditions in children. The research integrates clinical records, medical imaging, 3D facial phenotyping, speech and acoustic features, and behavioural data into unified diagnostic models. It spans the full pipeline from data harmonisation and model development through to clinician-in-the-loop validation and clinical translation, with potential for collaboration with clinical and research institutions domestically and internationally.
The candidate will be part of the digital health research team at the School of Electrical Engineering, Computing and Mathematical Sciences (EECMS), Curtin University, which is conducting this research in collaboration with cross-disciplinary researchers and global partners. The role sits at the intersection of machine learning, clinical informatics, and paediatric medicine, with opportunities to engage with clinical collaborators and contribute to international research in AI-enabled healthcare.
Aim
To develop a clinically validated multimodal AI framework that integrates heterogeneous paediatric health data for earlier detection of rare and neurodevelopmental conditions.
To design explainable AI methods that produce interpretable, clinician-facing outputs across imaging, speech, clinical text, and behavioural modalities.
To validate the framework across diverse patient cohorts and healthcare settings, assessing diagnostic accuracy, generalisability, and clinical usability.
To evaluate pathways for translating the framework into real-world clinical workflows, including decision support and diagnostic triage applications.
Objectives
To design and implement a multimodal data integration framework that harmonises heterogeneous paediatric clinical data, including electronic health records, imaging, speech, and behavioural assessments, into a unified and interoperable architecture suitable for AI modelling.
To develop deep learning models that capture diagnostic patterns across multiple clinical data modalities, using modality-specific encoders and multimodal fusion strategies to improve detection and stratification of rare and neurodevelopmental conditions.
To design and evaluate explainable AI methods that produce transparent, clinician-interpretable outputs, and assess their utility and usability through clinician-in-the-loop studies with paediatric specialists.
To validate the framework across diverse clinical datasets and healthcare settings, evaluate its integration into diagnostic workflows, and assess its potential as a scalable clinical decision support tool.
Significance
Diagnostic delays for rare and neurodevelopmental conditions routinely extend to five years or more. For children, this means missed windows for early intervention at the most developmentally sensitive periods of life, with significant costs to families and health systems.
Most AI diagnostic tools remain single-modality and single-site: they perform well on the data they were trained on and poorly on anything else. This project addresses the generalisability problem directly by designing for validation across diverse clinical environments and patient populations from the outset.
Explainability is treated here as a core design requirement, not an afterthought. A model that produces a diagnosis without explanation will not be adopted by clinicians regardless of its accuracy. Interpretability methods are embedded throughout the modelling pipeline and evaluated directly with clinical users.
The research contributes foundational advances in multimodal representation learning, cross-modality data fusion, and explainable AI for complex disease phenotypes. These methods have broad applicability beyond paediatric rare diseases, with potential translation into digital health platforms, clinical decision support systems, and population-level screening tools.
For the candidate, the project offers training across machine learning, clinical informatics, federated systems, and translational research, with a clear pathway to high-impact publications and a career at the interface of AI and healthcare.
Ideal Candidate
We are looking for a self-motivated PhD candidate with a strong degree in Computer Science, Software Engineering, or a closely related field, and demonstrated expertise in modern AI and machine learning. Strong communication skills and the ability to work effectively in a collaborative, cross-disciplinary team are essential. Some interest or background in health applications is an advantage. International applicants are expected to have at least one publication in an internationally recognised venue and must be able to cover their own tuition fees, as the RTP stipend does not automatically include a tuition fee offset for international students. Domestic applicants must demonstrate relevant research experience. Additionally, the applicants should meet the eligibility criteria for entry into a PhD program at Curtin University.
This project is open to International and Domestic applicants.
Internship
Through this project, you will also have the opportunity to undertake a paid internship with a sustainability industry partner, which is currently available to suitably qualified candidates.
Scholarship
If you are identified as the preferred candidate for this project, you may be considered for an RTP scholarship.
Enquires and How to Apply
For enquires about this opportunity contact Associate Professor Sonny Pham at Ducson.Pham@curtin.edu.au
To formally apply submit an Expression of Interest to Associate Professor Sonny Pham during the Central Scholarship round (July 1st – July 31st 2026)