FSU chemists develop AI tool for accurate chemical analysis

Chemists at Florida State University (FSU) have developed a new machine learning tool that can identify the chemical composition of dried salt solutions from images with 99% accuracy. This innovative tool combines robotics and artificial intelligence, making chemical analysis easier and more affordable. The researchers used a robot to prepare thousands of samples and an advanced machine learning program to analyze the data. Co-author Oliver Steinbock explained that by using large databases filled with images of different chemical compounds, AI can help determine their compositions. This advancement could lead to faster and cheaper chemical analyses in various fields, such as space exploration and law enforcement. In their previous work, the team analyzed 7,500 salt samples by hand. Now, with their Robotic Drop Imager (RODI), they can process over 2,000 samples a day. This automation allowed them to create a library of more than 23,000 images, significantly boosting the accuracy of their machine learning analysis. They simplified the images to grayscale and extracted key features to enhance the program's performance. The research showed the program could accurately identify salt types and concentrations, reaching up to 92% accuracy in distinguishing the different levels of salt solutions. Postdoctoral researcher Amrutha S.V. emphasized the benefits of a simple photographic method for chemical composition analysis, pointing out its potential applications in various fields. FSU’s efforts in artificial intelligence are generating new tools for research and teaching. The faculty are exploring innovative ways to use AI, which is transforming scientific discovery. Steinbock noted that this technology makes complex analyses more accessible, changing how research can be conducted in medicine and forensics, among other areas. The study was supported by NASA and involved additional researchers from Bowie State University and Florida A&M University.