Explore biomedical data
with your voice.

R. Adam, D. R. Catchpoole, S. J. Simoff, Z. Qu, P. J. Kennedy, Q. V. Nguyen.

Pacific Rim International Conference on Artificial Intelligence (PRICAI 2025), pp. 35-50. Springer Nature Singapore.

This paper introduces MediVerse, a system that integrates voice-based natural-language interfaces, immersive VR visualisation, and large-language-model query translation to enable real-time, hands-free interaction with complex biomedical datasets. The platform uses head-mounted VR displays for voice input, a cloud-based LLM for schema-aware query interpretation, and real-time 3D rendering of the resulting data. Evaluation across 20 seed queries on paediatric leukaemia (247 patients) and a de-identified paediatric tumour gene-expression cohort (101 samples) demonstrated accurate user-intent interpretation, low-latency responsiveness, and the correct selection of immersive visualisations in 95% of cases.

R. Adam, D. R. Catchpoole, S. J. Simoff, Z. Qu, P. J. Kennedy, Q. V. Nguyen.

Journal of Sensor and Actuator Networks, vol. 14, no. 2, art. 41 (2025).

The growing complexity and volume of genomic and omics data present critical challenges for storage, transfer, and analysis on edge-cloud platforms. This paper introduces a lossless compression method that integrates Trie-based shared dictionaries within an edge-cloud architecture. N-gram analysis identifies repeated sequences across entire patient cohorts; the resulting global dictionary is deployed to edge nodes for localised preprocessing, reducing redundancy before cloud transmission, where additional compression is then applied. Evaluation on B-ALL (178 patients, 7.92 GB) and CLL (162 patients, 3.98 GB) datasets achieved 68.2% reduction in file size compared with 57.9% for traditional ZSTD - an additional saving of approximately US $676 per year at this scale - while maintaining compatibility with standard bioinformatics pipelines.

R. Adam, D. R. Catchpoole, S. J. Simoff, P. J. Kennedy, Q. V. Nguyen.

Innovations in Digital Health, Diagnostics, and Biomarkers, vol. 4, pp. 81-88 (2024).

The healthcare landscape is rapidly integrating diverse data sources - electronic health records, omics, and genomic data - into unified patient profiles, enhancing personalised medicine and interoperability. This transformation, however, faces growing challenges in data integration and analysis. This study introduces a novel hybrid edge-cloud framework designed to manage the surge of multidimensional genomic and omics data in healthcare. It combines the localised processing of edge computing with the scalable resources of cloud computing, evaluated on simulated B-cell acute lymphoblastic leukaemia (B-ALL) and chronic lymphocytic leukaemia (CLL) cytometry datasets. The architecture achieved 56% data reduction, 57% faster upload times, enhanced privacy through local pseudonymisation, and projected annual cost savings of approximately US $4,679 on a 40 TB workload.

R. Adam, D. R. Catchpoole, S. J. Simoff, Z. Qu, P. J. Kennedy, Q. V. Nguyen.

Manuscript in preparation.

Existing evaluation practices for AI-driven immersive visualisation systems rely on static, single-axis metrics that fail to capture the joint quality of rendering, model decisions, and post-hoc explanation. IVEM (Immersive Visualisation Evaluation Model) introduces a unified metric suite of six operationalised metrics across three dimensions - Rendering Quality, AI Decision Quality, and Explanation Quality - validated on paediatric leukaemia and de-identified paediatric tumour gene-expression datasets. The suite enables diagnostic precision and objective cross-system comparison for AI-driven immersive analytics, complementing the MediVerse architecture introduced in earlier work.