Recently, the SEU Multidimensional Detection and Intelligent Recognition Team has made significant progress in the field of multi-energy X-ray imaging and intelligent substance identification.The related findings have been published in Science Advances, an international academic journal, under the title—Multi-energy X-ray imaging enabled by unipolar perovskite detector for intelligent substance identification. This research proposes a novel multi-energy imaging scheme based on a self-developed unipolar perovskite X-ray detection panel. This scheme eliminates the need for single-photon counting by relying on “voltage-encoded energy”, enabling automatic identification and pseudo-coloring of unknown substances.
This study developed a unipolar n–i–n perovskite X-ray detector, which demonstrates significantly stronger selective collection of electrons than holes. By modulating the internal electric field and carrier drift length with applied voltage, continuous-spectrum X-rays can be “decomposed” into distinct energy channels. Combined with a photo current decoupling algorithm under multiple voltages, the device can produce multi-energy images using a conventional X-ray source, thereby achieving high throughput, high sensitivity, and excellent spatial resolution.
On this basis, the research team further established a “multi-energy material database” and an intelligent recognition process for practical applications. First, absorption information from different energy channels was used to normalize the linear attenuation coefficient of each pixel so as to construct “energy fingerprints” composed of several σ(Ei)/σ(Ej) ratios, thereby mitigating the influence of sample thickness and dose fluctuations. Subsequently, standard sample data for common materials such as metals, polymers, and biological tissues were collected, and their energy fingerprint distributions were statistically analyzed to form a visualized material database. During imaging, the system matches the energy fingerprint of unknown pixels against the material database one by one. By integrating distance metrics and machine learning classification algorithms, automatic pixel-by-pixel identification and pseudo-coloring are achieved. This achievement enables different materials to be distinctly displayed and easily distinguished in X-ray images, even under conditions of complex overlapping layers and varying thicknesses.
The first author of the paper is Ph.D. student Li Yuwei from the School of Electronic Science & Engineering, SEU. The co-corresponding authors are Professor Xu Xiaobao and Professor Lei Wei from Southeast University, and Professor Gao Feng from Link?ping University, Sweden. This research was supported by the National Key Research and Development Program of China, the National Natural Science Foundation of China, and the Natural Science Foundation of Jiangsu Province, among other projects.
Paper'sPaper’s link: https://www.science.org/doi/10.1126/sciadv.adz0228
Source: School of Electronic Science and Engineering, SEU
Translated by: Melody Zhang
Proofread by: Gao Min
Edited by: Li Xinchang
