Supervised Laser-speckle Image Sampling of Skin Tissue to Detect Very Early Stage of Diabetes by Its Effects on Skin Subcellular Properties

Abstract

Objective: This paper investigates the effectiveness of a multi-disciplinary system based on laser speckle image sampling, image texture analysis, and artificial intelligence (AI), applied to the early detection of diabetes disease. The main objective is to develop an optimized unified technique to exploit the cellular form of skin structure and its low-level interaction with a specific laser light at critical wavelength which generates laser speckle effect to characterize sub-skin subtle cellular changes. This goal is achieved by the classification process which leads to detecting the presence of diabetes non-invasively from skin properties. Methods: With the latest developments in AI, laser-optics and laser speckle imaging technologies, it could be possible to optimize laser light parameters such as its wavelength, energy level, and its laser speckle image texture measures to achieve this goal. In the proposed work, such interdisciplinary techniques were demonstrated, which normally would not be possible by earlier conventional methods due to the lack of such optimized combinations. Results: Through the optimization of system parameters in laser-optic components, AI utilities, and texture analysis methods, it is possible to obtain up to 90% classification accuracy to identify normal and diabetes affected skin groups. Conclusion: The new approach is potentially more effective than the classical skin glucose level observation because of its optimized combination of laser physics and AI techniques. Additionally, it allows non-expert individuals to perform more frequent skin tissue tests for an early detection of diabetes. Cost effectiveness, which is another target for the proposed system, has also been justified by using low-cost system components, such as a very low-power laser source and normal digital camera.