CoChat
Sign Up
Cited by 110Radiation Detection and Scintillator TechnologiesAdvanced Semiconductor Detectors and MaterialsPerovskite Materials and Applications

The DOI matches its Crossref record.

Reconfigurable perovskite X-ray detector for intelligent imaging

Open access
Pang, Jincong, Wu, Haodi, Li, Hao, Jin, Tong, Tang, Jiang, Niu, Guangda·2024·Nature Communications
DOI 10.1038/s41467-024-46184-0PMC PMC10899650
Open this paper in CoChat — free accountOpen PDF

Abstract

X-ray detection is widely used in various applications. However, to meet the demand for high image quality and high accuracy diagnosis, the raw data increases and imposes challenges for conventional X-ray detection hardware regarding data transmission and power consumption. To tackle these issues, we present a scheme of in-X-ray-detector computing based on CsPbBr 3 single-crystal detector with convenient polarity reconfigurability, good linear dynamic range, and robust stability. The detector features a stable trap-free device structure and achieves a high linear dynamic range of 106 dB. As a result, the detector could achieve edge extraction imaging with a data compression ratio of ~50%, and could also be programmed and trained to perform pattern recognition tasks with a high accuracy of 100%. Our research shows that in-X-ray-detector computing can be used in flexible and complex scenarios, making it a promising platform for intelligent X-ray imaging.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41467-024-46184-0

Cites

Statistics

Figures

Fig. 1 Working mechanism of in-X-ray-detector computing. a Schematic diagram of the in-X-ray-detector computing using macro- and sub-pixel. It uses the Laplacian convolution kernel, and the weight setting of the kernel is implemented using the bias voltages of the sub-pixel. Pictures of packages containing various metal tools are used to demonstrate the effect and results of convolution kernels. The three pictures mean that a huge amount of raw data is obtained without image processing, less data is obtained with the LoG convolution kernel, and little data is obtained with the Laplacian kernel. W/O represents without, LoG represents Laplacian of the Gaussian. Baggages-0002-0001 is from the publicly available database, GDXray+50. b The development trends of the maximum amount of data detected per second and the transmission rates in the computed tomography (CT) field. The former is represented by red patterns and lines. The latter is represented by blue patterns and lines. Among them, products from United Image are represented by circles, products from General Electric Company (GE) are represented by squares, products from Siemens are represented by triangles, products from Toshiba are represented by rhombus, products from Philips are represented by stars, and products from Neusoft are represented by pentagons. Schleifring and Moog’s products use rectangles to delineate ranges of rates, and dots to represent typical values. The orange line is the compressed detected data rate after implementing the convolutional kernel in this work. Details can be found in Supplementary Note 1, and Tables 1 and 2. c The bias-tunable X-ray response with the same magnitude of perovskite detectors. This is the premise of the in-X-ray-detector computing architecture.
Fig. 1 Working mechanism of in-X-ray-detector computing. a Schematic diagram of the in-X-ray-detector computing using macro- and sub-pixel. It uses the Laplacian convolution kernel, and the weight setting of the kernel is implemented using the bias voltages of the sub-pixel. Pictures of packages containing various metal tools are used to demonstrate the effect and results of convolution kernels. The three pictures mean that a huge amount of raw data is obtained without image processing, less data is obtained with the LoG convolution kernel, and little data is obtained with the Laplacian kernel. W/O represents without, LoG represents Laplacian of the Gaussian. Baggages-0002-0001 is from the publicly available database, GDXray+50. b The development trends of the maximum amount of data detected per second and the transmission rates in the computed tomography (CT) field. The former is represented by red patterns and lines. The latter is represented by blue patterns and lines. Among them, products from United Image are represented by circles, products from General Electric Company (GE) are represented by squares, products from Siemens are represented by triangles, products from Toshiba are represented by rhombus, products from Philips are represented by stars, and products from Neusoft are represented by pentagons. Schleifring and Moog’s products use rectangles to delineate ranges of rates, and dots to represent typical values. The orange line is the compressed detected data rate after implementing the convolutional kernel in this work. Details can be found in Supplementary Note 1, and Tables 1 and 2. c The bias-tunable X-ray response with the same magnitude of perovskite detectors. This is the premise of the in-X-ray-detector computing architecture.CC BY · PubMed Central
Fig. 2 Study of defects in PSC with and without the C60 passivation layer. a X-ray photoelectron spectroscopy (XPS) and photoluminescence (PL) spectra. b Thermally stimulated current (TSC) curves. The results of split-peak fitting can be viewed in Supplementary Fig. 4 and the method can be viewed in Supplementary Note 2. c Schematic diagram of photoconductive gain effect with and without defect energy levels. The defect level traps one type of photogenerated carrier. To maintain the neutrality of the semiconductor, the electrodes inject carriers. This will result in the photoconductive gain. d, e I–V and I–t curve. w/o represents without. f Simulation values of the intrinsic sensitivity with various thicknesses and μτE for the perovskite single crystal (PSC). Details can be found in Supplementary Note 3. Source data are provided as a Source Data file.
Fig. 2 Study of defects in PSC with and without the C60 passivation layer. a X-ray photoelectron spectroscopy (XPS) and photoluminescence (PL) spectra. b Thermally stimulated current (TSC) curves. The results of split-peak fitting can be viewed in Supplementary Fig. 4 and the method can be viewed in Supplementary Note 2. c Schematic diagram of photoconductive gain effect with and without defect energy levels. The defect level traps one type of photogenerated carrier. To maintain the neutrality of the semiconductor, the electrodes inject carriers. This will result in the photoconductive gain. d, e I–V and I–t curve. w/o represents without. f Simulation values of the intrinsic sensitivity with various thicknesses and μτE for the perovskite single crystal (PSC). Details can be found in Supplementary Note 3. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 3 The perovskite single-crystal device performance. a μτ fitted by the Hecht formula. b, c The I–V and I–t curves under forward and reverse bias voltages with variable X-ray intensity. The illustration shows the N-I-P perovskite single-crystal (PSC) device structure and the energy bands are obtained according to Supplementary Fig. 1. d Linear dynamic range (LDR) results. e Comparison of LDR of semiconductor X-ray detectors. f The area array detector integrated onto the complementary metal-oxide-semiconductor (CMOS), the optical photo, and the structural diagram of the CMOS die. Detailed schematic diagrams and process photos can be viewed in Supplementary Figs. 8 and 9. CDS represents correlated double sampling, PGA represents programmable gain amplifier, and DI represents direct injection. Source data are provided as a Source Data file.
Fig. 3 The perovskite single-crystal device performance. a μτ fitted by the Hecht formula. b, c The I–V and I–t curves under forward and reverse bias voltages with variable X-ray intensity. The illustration shows the N-I-P perovskite single-crystal (PSC) device structure and the energy bands are obtained according to Supplementary Fig. 1. d Linear dynamic range (LDR) results. e Comparison of LDR of semiconductor X-ray detectors. f The area array detector integrated onto the complementary metal-oxide-semiconductor (CMOS), the optical photo, and the structural diagram of the CMOS die. Detailed schematic diagrams and process photos can be viewed in Supplementary Figs. 8 and 9. CDS represents correlated double sampling, PGA represents programmable gain amplifier, and DI represents direct injection. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 4 Imaging for edge extraction. a The schematic diagram of imaging. b The X-ray imaging of the experiment, following conventional methods. c The imaging following the in-X-ray-detector computing method (the weight values of the pixels are allocated according to the Laplacian convolution kernel). d The gray distribution histogram of (c). e The histogram of the gray distribution of Baggages-0002-0001 with (red) or without (blue) Laplacian simulation processing. In the blue grayscale histogram, which represents the result processed by the convolution kernel, 686,739 pixels have a grayscale value less than 1. f–i Other convolution kernels with their pixel weights, and the effects of the simulation image stylization. All scale bars in red color are 1 cm. Source data are provided as a Source Data file.
Fig. 4 Imaging for edge extraction. a The schematic diagram of imaging. b The X-ray imaging of the experiment, following conventional methods. c The imaging following the in-X-ray-detector computing method (the weight values of the pixels are allocated according to the Laplacian convolution kernel). d The gray distribution histogram of (c). e The histogram of the gray distribution of Baggages-0002-0001 with (red) or without (blue) Laplacian simulation processing. In the blue grayscale histogram, which represents the result processed by the convolution kernel, 686,739 pixels have a grayscale value less than 1. f–i Other convolution kernels with their pixel weights, and the effects of the simulation image stylization. All scale bars in red color are 1 cm. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 5 Implementation of a convolutional neural network. a Training process diagram. PC, personal computer. b, c Some objects and their detected X-ray pictures are used in the database. All scale bars in red color are 1 cm. d Accuracy of recognition over the epochs. e Output values, fH, fU, fS, and fT, for each epoch. The four boxplots, the y-axis of which are fH, fU, fS, and fT, respectively, from top to bottom, represent the processing results of the four feature convolution kernels. The grid lines of the box are divided according to 25%, 50%, and 75% of the data. Percentiles are represented by horizontal lines for minimum and maximum values, crosses for 1% and 99%, and squares for mean values. The colors black, red, blue, and yellow represent the letters H, U, S, and T, respectively. Source data are provided as a Source Data file.
Fig. 5 Implementation of a convolutional neural network. a Training process diagram. PC, personal computer. b, c Some objects and their detected X-ray pictures are used in the database. All scale bars in red color are 1 cm. d Accuracy of recognition over the epochs. e Output values, fH, fU, fS, and fT, for each epoch. The four boxplots, the y-axis of which are fH, fU, fS, and fT, respectively, from top to bottom, represent the processing results of the four feature convolution kernels. The grid lines of the box are divided according to 25%, 50%, and 75% of the data. Percentiles are represented by horizontal lines for minimum and maximum values, crosses for 1% and 99%, and squares for mean values. The colors black, red, blue, and yellow represent the letters H, U, S, and T, respectively. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

No references could be resolved yet.

Research works printed but not linked: 52

  1. Spahn, M. X-ray detectors in medical imaging. Nucl. Instrum. Meth. A 731, 57–63 (2013). · DOI
  2. Pang, J. et al. Vertical matrix perovskite X-ray detector for effective multi-energy discrimination. Light Sci. Appl. 11, 105 (2022). · DOI
  3. Duan, X. et al. X-ray cargo container inspection system with few-view projection imaging. Nucl. Instrum. Meth. A 598, 439–444 (2009). · DOI
  4. Haff, R. P. & Toyofuku, N. X-ray detection of defects and contaminants in the food industry. Sens. Instrum. Food Qual. 2, 262–273 (2008). · DOI
  5. Neutze, R., Wouts, R., Van der Spoel, D., Weckert, E. & Hajdu, J. Potential for biomolecular imaging with femtosecond X-ray pulses. Nature 406, 752–757 (2000). · DOI · Paper page
  6. Thanki, R. M. & Ashish K. Hybrid and Advanced Compression Techniques for Medical Images (Springer International Publishing, 2019) · DOI
  7. Patidar, G., Kumar, S. & Kumar, D. A review on medical image data compression techniques. In: 2nd Int. Conf. Data, Eng. Appl. (IDEA), 1–6 (IEEE, 2020) · DOI
  8. Hammer, M., Yoshii, K. & Miceli, A. Strategies for on-chip digital data compression for x-ray pixel detectors. J. Instrum. 16, P01025 (2021). · DOI
  9. Strempfer, S. et al. A lightweight, user-configurable detector ASIC digital architecture with on-chip data compression for MHz X-ray coherent diffraction imaging. J. Instrum. 17, P10042 (2022). · DOI
  10. Barman, R., Badade, S., Deshpande, S., Agarwal, S. & Kulkarni, N. Lossless data compression method using deep learning. In: Mach. Intell. Smart Syst. 145–151 (MISS, 2022) · DOI
  11. Pi, L. et al. Broadband convolutional processing using band-alignment-tunable heterostructures. Nat. Electron. 5, 248–254 (2022). · DOI
  12. Wang, C. Y. et al. Gate-tunable van der Waals heterostructure for reconfigurable neural network vision sensor. Sci. Adv. 6, eaba6173 (2020). · DOI
  13. Jayachandran, D. et al. A low-power biomimetic collision detector based on an in-memory molybdenum disulfide photodetector. Nat. Electron. 3, 646–655 (2020). · DOI
  14. Moin, A. et al. A wearable biosensing system with in-sensor adaptive machine learning for hand gesture recognition. Nat. Electron. 4, 54–63 (2021). · DOI
  15. Kasap, S. O., Koughia, K. V., Fogal, B., Belev, G. & Johanson, R. E. The influence of deposition conditions and alloying on the electronic properties of amorphous selenium. Semiconductors 37, 789–794 (2003). · DOI
  16. Owens, A. Semiconductor materials and radiation detection. J. Synchrotron Radiat. 13, 143–150 (2006). · DOI
  17. Zhou, Y. et al. Self-powered perovskite photon-counting detectors. Nature 616, 1–7 (2023). · DOI
  18. Buonanno, L., Di Vita, D., Carminati, M. & Fiorini, C. Gamma: a 16-channel spectroscopic asic for sipms readout with 84-db dynamic range. IEEE Trans. Nucl. Sci. 68, 2559–2572 (2021). · DOI
  19. Zhuang, R. et al. Highly sensitive X-ray detector made of layered perovskite-like (NH4)3Bi2I9 single crystal with anisotropic response. Nat. Photon. 13, 602–608 (2019). · DOI
  20. Zhang, Y. et al. Nucleation-controlled growth of superior lead-free perovskite Cs3Bi2I9 single-crystals for high-performance X-ray detection. Nat. Commun. 11, 2304 (2020). · DOI
  21. Li, H. et al. Sensitive and stable 2D perovskite single-crystal X-ray detectors enabled by a supramolecular anchor. Adv. Mater. 32, 2003790 (2020). · DOI
  22. Wei, H. et al. Dopant compensation in alloyed CH3NH3PbBr3-xClx perovskite single crystals for gamma-ray spectroscopy. Nat. Mater. 16, 826–833 (2017). · DOI
  23. Lai, P. T. et al. All-Vacuum-deposited perovskite X-ray detector with a record-high self-powered sensitivity of 1.2 C Gy-1 cm−3. ACS Appl. Mater. Interfaces 14, 19795–19805 (2022). · DOI
  24. Guo, F. et al. A nanocomposite ultraviolet photodetector based on interfacial trap-controlled charge injection. Nat. Nanotechnol. 7, 798–802 (2012). · DOI
  25. Konstantatos, G., Clifford, J., Levina, L. & Sargent, E. H. Sensitive solution-processed visible-wavelength photodetectors. Nat. Photon. 1, 531–534 (2007). · DOI
  26. Fang, Y., Armin, A., Meredith, P. & Huang, J. Accurate characterization of next-generation thin-film photodetectors. Nat. Photon. 13, 1–4 (2019). · DOI
  27. Liu, D. et al. Impact of ultrathin C60 on perovskite photovoltaic devices. ACS Nano 12, 876–883 (2018). · DOI
  28. Muscarella, L. et al. Trap passivation and suppressed electrochemical dynamics in perovskite solar cells with C60 interlayers. Electrochim. Acta 433, 141215 (2022). · DOI
  29. Wojciechowski, K. et al. C60 as an efficient n-type compact layer in perovskite solar cells. J. Phys. Chem. Lett. 6, 2399–2405 (2015). · DOI
  30. Liu, X. et al. Novel efficient C60-based inverted perovskite solar cells with negligible hysteresis. Electrochim. Acta 288, 115–125 (2018). · DOI
  31. Zhang, M. et al. Determination of defect levels in melt-grown all-inorganic perovskite CsPbBr3 crystals by thermally stimulated current spectra. J. Phys. Chem. C. 122, 10309–10315 (2018). · DOI
  32. He, Y. et al. High spectral resolution of gamma-rays at room temperature by perovskite CsPbBr3 single crystals. Nat. Commun. 9, 1609 (2018). · DOI
  33. Wang, F. et al. Precursor engineering for solution method-grown spectroscopy-grade CsPbBr3 crystals with high energy resolution. Chem. Mater. 34, 3993–4000 (2022). · DOI
  34. Ni, Z. et al. Resolving spatial and energetic distributions of trap states in metal halide perovskite solar cells. Science 367, 1352–1358 (2020). · DOI
  35. Fei, C. et al. Lead-chelating hole-transport layers for efficient and stable perovskite minimodules. Science 380, 823–829 (2023). · DOI
  36. Kang, J. & Wang, L. W. High defect tolerance in lead halide perovskite CsPbBr3. J. Phys. Chem. Lett. 8, 489–493 (2017). · DOI
  37. Di, J. et al. Reveal the humidity effect on the phase pure CsPbBr3 single crystals formation at room temperature and its application for ultrahigh sensitive X-ray detector. Adv. Sci. 9, 2103482 (2022). · DOI
  38. Peng, J. et al. Crystallization of CsPbBr3 single crystals in water for X-ray detection. Nat. Commun. 12, 1531 (2021). · DOI
  39. Zhang, P. et al. Ultrasensitive and robust 120 keV hard X-Ray imaging detector based on mixed-halide perovskite CsPbBr3-nIn single crystals. Adv. Mater. 34, 2106562 (2022). · DOI
  40. Panneerselvam, D. M. & Kabir, M. Z. Evaluation of organic perovskite photoconductors for direct conversion X-ray imaging detectors. J. Mater. Sci. Mater. Electron. 28, 7083–7090 (2017). · DOI
  41. Greiffenberg, D., Fauler, A., Zwerger, A. & Fiederle, M. Energy resolution and transport properties of CdTe-Timepix-Assemblies. J. Instrum. 6, C01058 (2011). · DOI
  42. Toufanian, R., Swain, S., Becla, P., Motakef, S. & Datta, A. Cesium lead bromide semiconductor radiation detectors: crystal growth, detector performance and polarization. J. Mater. Chem. C. 10, 12708–12714 (2022). · DOI
  43. Tsai, H. et al. A sensitive and robust thin-film x-ray detector using 2D layered perovskite diodes. Sci. Adv. 6, eaay0815 (2020). · DOI
  44. Xu, X. et al. Sequential growth of 2D/3D double-layer perovskite films with superior X-ray detection performance. Adv. Sci. 8, 2102730 (2021). · DOI
  45. Wei, W. et al. Monolithic integration of hybrid perovskite single crystals with heterogenous substrate for highly sensitive X-ray imaging. Nat. Photon. 11, 315–321 (2017). · DOI
  46. Pan, L. et al. Ultra-high flux X-ray detection by solution-grown perovskite CsPbBr3 single crystal semiconductor detector. Adv. Mater. 35, 2211840 (2023). · DOI
  47. He, Y. et al. Sensitivity and detection limit of spectroscopic-grade perovskite CsPbBr3 crystal for hard X-ray detection. Adv. Funct. Mater. 32, 2112925 (2022). · DOI
  48. Kim, Y. et al. Printable organometallic perovskite enables large-area, low-dose X-ray imaging. Nature 550, 87–91 (2017). · DOI
  49. Deumel, S. et al. High-sensitivity high-resolution X-ray imaging with soft-sintered metal halide perovskites. Nat. Electron. 4, 681–688 (2021). · DOI
  50. Mery, D. et al. GDXray: the database of X-ray images for nondestructive testing. J. Nondestruct. Eval. 34, 1–12 (2015). · DOI
  51. Gatys, L. A., Ecker, A. S. & Bethge, M. Image style transfer using convolutional neural networks. In: Proc. IEEE Conf. Comput. Vis. Pattern Recognit. 2414–2423 (IEEE, 2016) · DOI · Paper page
  52. Jang, H. et al. In-sensor optoelectronic computing using electrostatically doped silicon. Nat. Electron. 5, 519–525 (2022). · DOI

Cited by

Sorted by citations

Metadata

Authors
Pang, J., Wu, H., Li, H., Jin, T., Tang, J., & Niu, G.
Affiliations
Wuhan National Laboratory for Optoelectronics (CN)
Corresponding author
Niu, Guangda
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2024-02-27
Volume
15
Issue
1
Pages
1769-1769
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-024-46184-0
PMC
PMC10899650
OpenAlex
W4392200309
Cited by
110
References
0
Citations by year
2026: 35 · 2025: 50 · 2024: 25
FWCI
32.57
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Natural Science Foundation of China (62134003, 12050005, 62074066)
  • Fundamental Research Funds for the Central Universities
Topics
  • Physical Sciences › Physics and Astronomy › Radiation › Radiation Detection and Scintillator Technologies
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advanced Semiconductor Detectors and Materials
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Perovskite Materials and Applications
Keywords
in-sensor computing, linear dynamic range, radiography, edge detection, data compression, pattern recognition

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review