Image processing, color detection, and supervised Machine Learning models in silk-based and thermo-colorimetric sensors

In the biomedical field, the use of machine learning and computer vision analysis is growing rapidly, enabling better devices and meaningful analyses that help solve many problems in research topics and concrete applications. The need to use biomaterials, integrated with advanced technologies, is becoming an opportunity to create sustainable and biocompatible opportunities for the technological world. Living materials can be considered an interface between life science and technology. Silk is a very powerful material. It is a natural fiber that contains natural fibers such as fibroin, which are useful for creating interesting materials with special characteristics. Silklab is one of the world's most specialized laboratories for developing biomaterials (silk-based and otherwise). It is located in Boston (MA, USA) and is affiliated with Tufts University. In 2022, several projects were aimed at combining machine learning algorithms and computer vision analysis to develop mechanical and thermo-colorimetric sensors, and to analyze some living material tissues. Producing sensors without an electronic component and associating some form of predictive modeling with them allows the use of bio-edible, bio-compatible devices with a wide applicability. Moreover, through image analysis, color detection, and image recognition, some of these sensors can be functionally studied to achieve remarkable results. Some projects have also been tested to animal tissues, such as octopus skin, which, when analyzed under a microscope, showed interesting patterns for further applications. Not only silk-based patches, but also ink sensors sensitive to temperature have been designed and used to develop sensors for human health. These studies can be significant in paving the way for a world of reusable and accessible devices. In this master's thesis, the first part focuses on the analysis of textiles and thermo-colorimetric inks to understand their behavior and find the best suited way to analyze them. Then, part of these results and methods were applied to design and create other types of sensors. Specifically, patched sensors to detect the amount of lactate produced by humans after physical activities and thermo-colorimetric sensors to detect the temperatures of objects or living things. In these studies, supervised machine learning techniques have been applied to achieve certain goals, such as finding the amount of lactate or accurately detecting the temperature of a specific body. In summary, the project described in this thesis investigates computer vision analysis and predictive models to intersect and mix bio-reality with the AI world, producing an interaction of heterogeneous knowledge aimed at developing a technology from the planet, for the planet.