YuWei Li, Xiangyu Ou, Shilin Liu, Jingda Zhao, Lu Xue, Chengjun Liu, Bingjian Zhu, Su Yan, S.D. Wang, Ziyu Wei, Sajid Hussain, Wei Lei, Feng Gao, Xiaobao Xu
A challenge with state-of-the-art projection x-ray imaging technologies is their limited ability to identify unknown substances. Here, we develop an intelligent multienergy x-ray imaging technique capable of precisely distinguishing different substances and labeling them with diverse colors. Our design uses a series of x-ray attenuation coefficient ratios under different x-ray energies as substance-specific markers. For this purpose, unipolar perovskite x-ray detectors are carefully engineered to resolve the x-ray energies into seven channels using a customized algorithm. Combining machine learning and a comprehensive x-ray attenuation ratio database of common materials enables accurate recognition of low-density biological tissues composed of light elements with similar atomic numbers. By transforming the intensity scale in conventional x-ray images into an attenuation coefficient ratio, our work presents a proof of concept for color-coded x-ray imaging, highlighting its potential for applications in energy-dispersive computed tomography, targeted drug delivery, quantum physics, and universe exploration.