One of the pillars of the Digital Patient technologies will be that of "information blending, or fusion". We take a look at what this might mean in practise.
The ability to process, manipulate and combine health information already plays a big part in clinical practise, and this ability will increase enormously with Digital Patient technologies. Currently, medical imaging is routinely used to diagnose and understand pathologies. The raw imaging data provides useful information itself, but the use of image analysis technology allows us to extract far richer knowledge. Today, this includes image segmentation, allowing us to separate tissue types and visualise specific structures in 3D. Image registration can help us to map different datasets together, and extract useful combined information.
Meanwhile, physiological models can provide predictive tools based on the underlying physiological and biological processes. These two technologies – imaging analysis and physiological modelling – can be combined in ways which can produce whole new types of information. For example, mechanical analysis such as Finite Element Analysis can be used to quantify mechanical stimuli in tissue using medical imaging as an input. This can then be mapped back to the medical image as a new‘in silico’-generated image1, showing patterns of stress in the tissue.
The integration of models and imaging extends to validation, for example when fluid flow models of cardiovascular fluid flow are compared to angiogram data. In future, this integration could go even further, with new technologies for information fusion. One interesting example is in multi-modal image registration. In this type of analysis, image information from two modalities (a chest MRI and mammography, for breast cancer imaging, for example) is combined in order to generate more useful information1. A key problem is that the tissue shape may vary considerably for each modality – depending on the patient and tissue position. In this case, mechanical models of tissue deformation can help register the images by simulating the deformations the tissue undergoes.
This use of models to enrich an existing information dataset has the potential to produce novel and valuable new types of information.