The Impact of Hypercube Compression on the Diagnostic Performance of a Hyperspectral Imaging-Based Telemedicine System

<p dir="ltr">The United Nations Sustainable Development Goal 3 of ensuring good health and well-being for all by 2030 is significantly impeded by the limited availability of quality specialist healthcare services in rural, remote, and underserved communities. Telemedicine offers a potential solution by facilitating the provision of specialist healthcare services to patients in rural communities from urban healthcare centres via the Internet. Hyperspectral imaging (HSI) is an emerging medical imaging modality that enables specialist healthcare applications like skin cancer diagnosis and, potentially, latent tuberculosis screening. An HSI-based telemedicine system was developed for rural, remote, and underserved communities, that spatially compresses hyperspectral images (also called hypercubes) based on the mobile network quality of a rural healthcare centre before transmitting them to a city-based healthcare centre server. Its limitation is the lack of a diagnostic component which would have facilitated the investigation of the effect of spatial compression on HSI-based diagnosis. Thus, in this paper, we address this research gap by updating the HSI-based telemedicine architecture with a deep learning-based diagnostic model to investigate the impact of medical hypercube compression on diagnostic performance metrics. The updated telemedicine system was updated using an opensource dataset of skin cancer hypercubes. Our results show the discriminatory ability of the HSI-based telemedicine system to distinguish positive from negative state of a disease to be constant for hypercubes compressed up to 45% of their spatial dimensions which results in a 70% reduction in hypercube file size. With these results we demonstrate the ability of an HSI-based telemedicine architecture to provide remote diagnosis while demonstrating its ability to maximise limited network resources by minimising hypercube file size required for consistent diagnosis. This further strengthens the evidence supporting the capacity of telemedicine to alleviate the scarcity of specialist healthcare services in rural, remote, and underserved communities.</p>