Efficient Mapping of Tissue Oxygen Saturation Using Hyperspectral Imaging and GAN

Abstract

Oxygen saturation is a critical parameter widely used in clinical practice to assess the concentration of oxygen in the blood. Recently, hyperspectral imaging-based oximetry techniques have been proposed for broad applications of two-dimensional information without sensor contact. However, these methods suffer from high computational complexity and inefficiency due to the large size of hyperspectral data cubes. In this study, we investigate an efficient tissue oxygen saturation (StO2) mapping from hyperspectral image using a generative adversarial network (GAN). The model was trained on hyperspectral images acquired from 40 healthy individuals. To reduce computational load while maintaining performance, the Non-dominated Sorting Genetic Algorithm II (NSGA-II) was then employed to identify the optimal combination of hyperspectral bands. The selected optimal band combinations included both visible and near-infrared (NIR) bands. The GAN-based model with the ten-band combination estimated StO2 effectively, achieving a root mean squared error (RMSE) of $0.1169~\pm ~0.0350$ . The trained model was then applied to palm and sole images of 20 patients with type 2 diabetes mellitus (DM) to validate its performance. The performance on patient data was comparable to that obtained from healthy individuals, with the best results achieved using the ten-band combination (RMSE $= 0.1077~\pm ~0.0341$ ). In conclusion, the developed efficient hyperspectral image-based oxygenation mapping method demonstrated robust performance for patients with type 2 DM as well as healthy subjects.

Keywords

• Tissue oxygen saturation
• hyperspectral imaging
• generative adversarial network
• hyperspectral band selection
• type 2 diabetes mellitus