Generative AI Enhanced Hyperspectral Imaging Techniques for Estimating Bone Fracture Risk in Osteoporosis

Osteoporosis (OP) is a global burden affecting 3.5m individuals in the UK alone, causing bone weakening and fractures ranking the fourth greatest cause of death. Bone Mineral Density (BMD) via dual- energy X-ray absorptiometry (DXA) currently serves as the tool to diagnose OP and risk of fractures. However, BMD is insufficient to fully describe bone tissues quality so mechanical properties must be considered.

Hyperspectral Imaging (HSI) is a non-invasive and quick imaging technique that offers valuable diagnostic insights. In recent years, medical HSI has shown promising outcomes across various medical fields, such as oncology, digital and computational pathology, ophthalmology, dermatology, and gastroenterology. It captures spatial and spectral information, making it possible to identify and characterize different tissues through their unique spectral signatures. By taking images across multiple spectral bands at the same time, HSI can detect abnormal tissue characteristics that traditional imaging methods might miss. Therefore, HSI can complement and enhance the diagnostic capabilities of other imaging methods.

In this project, our aim is to create a non-invasive, mechanically informed approach, calibrating HSI on X-ray tomography data. This will enable us to assess bone tissue quality in detail and identify potential biomarkers for osteoporosis-related fractures.

This project builds on our preliminary generative AI method (GenAI) [1]. It leverages GenAI’s capabilities to enhance tissue analysis in HSI data, facilitating the early identification of bone fracture risks or symptoms not visible in the initial stages of diseases such as osteoporosis. This enables timely intervention and treatment.

We expect the new algorithm to achieve a significant result on the new bone HSI dataset, improving bone fracture prediction.

This project is a collaborative effort between the School of Computing & Mathematical Sciences (CMS) and the School of Engineering (SoE) to utilise expertise and facilities between two schools.