Christopher J. Fluke

Belal Alsinglawi, Chris McCarthy, Sara Webb et al.

Medical Visual Question Answering (VQA) enhances clinical decision-making by enabling systems to interpret medical images and answer clinical queries. However, developing efficient, high-performance VQA models is challenging due to the complexity of medical imagery and diverse modalities. In this paper, we introduce a lightweight, multimodal VQA model integrating BiomedCLIP for image feature extraction and LLaMA-3 for text processing. Designed for medical VQA tasks, our model achieves state-of-the-art performance on the OmniMedVQA dataset. With approximately 8 billion parameters, it requires only two NVIDIA 40 GB A100 GPUs, demonstrating superior efficiency over larger models. Our results show 73.4% accuracy for open-end questions, surpassing existing models and validating its potential for real-world medical applications. Key contributions include a specialized multimodal VQA model, a resource-efficient architecture, and strong performance in answering open-ended clinical questions.

Dany Vohl, David G. Barnes, Christopher J. Fluke et al.

We present encube $-$ a qualitative, quantitative and comparative visualisation and analysis system, with application to high-resolution, immersive three-dimensional environments and desktop displays. encube extends previous comparative visualisation systems by considering: 1) the integration of comparative visualisation and analysis into a unified system; 2) the documentation of the discovery process; and 3) an approach that enables scientists to continue the research process once back at their desktop. Our solution enables tablets, smartphones or laptops to be used as interaction units for manipulating, organising, and querying data. We highlight the modularity of encube, allowing additional functionalities to be included as required. Additionally, our approach supports a high level of collaboration within the physical environment. We show how our implementation of encube operates in a large-scale, hybrid visualisation and supercomputing environment using the CAVE2 at Monash University, and on a local desktop, making it a versatile solution. We discuss how our approach can help accelerate the discovery rate in a variety of research scenarios.

Bernard F Meade, Christopher J Fluke, Steven Manos et al.

Clustering commodity displays into a Tiled Display Wall (TDW) provides a cost-effective way to create an extremely high resolution display, capable of approaching the image sizes now gen- erated by modern astronomical instruments. Astronomers face the challenge of inspecting single large images, many similar images simultaneously, and heterogeneous but related content. Many research institutions have constructed TDWs on the basis that they will improve the scientific outcomes of astronomical imagery. We test this concept by presenting sample images to astronomers and non- astronomers using a standard desktop display (SDD) and a TDW. These samples include standard English words, wide field galaxy surveys and nebulae mosaics from the Hubble telescope. These experiments show that TDWs provide a better environment for searching for small targets in large images than SDDs. It also shows that astronomers tend to be better at searching images for targets than non-astronomers, both groups are generally better when employing physical navigation as opposed to virtual navigation, and that the combination of two non-astronomers using a TDW rivals the experience of a single astronomer. However, there is also a large distribution in aptitude amongst the participants and the nature of the content also plays a significant role is success.