Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Jun 10;14(1):13256.
doi: 10.1038/s41598-024-63834-x.

Electrospray deposition of physical unclonable functions for drug anti-counterfeiting

Bryce J Kingsley  1 , J David Schaffer  2 , Paul R Chiarot  3
Affiliations

Electrospray deposition of physical unclonable functions for drug anti-counterfeiting

Bryce J Kingsley et al. Sci Rep. .

Abstract

In recent years, pharmaceutical counterfeiting has become an increasingly dangerous situation. A patient who unknowingly consumes a counterfeit drug is at a serious health risk. To address this problem, a low-cost and robust approach for authentication that can be administered at the point-of-care is required. Our proposed solution uses Optical Physical Unclonable Functions (PUFs); patterns formed by a stochastic process that can be used for authentication. We create edible PUFs (ePUFs) using electrospray deposition, which utilizes strong electric fields to atomize a liquid suspension into a plume of micro-scale droplets that are delivered to the target. The ePUFs are electrospray-deposited from an edible ink directly onto the surface of the drug tablets. The process parameters (flow rate, translation speed, and suspension concentration) govern the characteristics of the ePUF to provide highly stochastic patterns. To evaluate our approach, 200 ePUFs were deposited onto tablets at various conditions, followed by imaging and storage of the patterns in a database. For ePUF authentication, a machine vision approach was created using the open source SIFT pattern matching algorithm. Using optimized pattern-matching constraints, our algorithm was shown to be 100% successful in authenticating the cellphone images of the ePUFs to the database. Additionally, the algorithm was found to be robust against changes in illumination and orientation of the cellphone images.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1
Figure 1
Schematic of electrospray-deposited edible-PUF (ePUF) onto drug tablets for per-dose authentication. ePUFs are deposited onto the pills using electrospray. Standardized images of the ePUFs are captured and stored in a database before the pill is shipped to the patient. The patient captures an image with their cellphone camera which is checked against the database to authenticate the pill.
Figure 2
Figure 2
Schematic of ePUF deposition onto a pill. High voltage is applied to the edible ink solution to generate the electrospray. A plastic stencil is placed on top of the target pill, which is translated in a zig-zag pattern to deposit the ePUF on its surface through the opening in the stencil.
Figure 3
Figure 3
Process flow diagram for ePUF matching.
Figure 4
Figure 4
Image processing steps to extract ePUF pattern from an image of the pill. The raw image (1) is cropped to the pill and the background is removed (2), from which the ePUF pattern is extracted (3).
Figure 5
Figure 5
Effects of solution flow rate (left), translation speed (middle), and mass loading (right) on deposited ePUF pattern. Solution flow rate and translation speed affect the pattern density, and mass loading alters the color intensity of the pattern.
Figure 6
Figure 6
The effect of distance ratio k on ePUF pattern matching. Plots (a) and (b) show the results of pattern matching without and with symmetry checking, respectively. Colored bars represent different values of k, as specified in the legends. A prominence threshold of p=0.5 was used as the confidence criterion with a matching threshold λ=0.02.
Figure 7
Figure 7
Representative images of correct ePUF matches between cellphone and database images. Top left (smaller) images show the ePUFs extracted from cellphone images. The larger images are the ePUFs from the database. The smaller (cellphone) image is rotated to match the orientation of the larger (database) image. The colored points show the SIFT features (keypoints) in each image, and the lines represent good feature matches between the images. Panel (a) shows the results for a less dense ePUF (pattern #93) and (b) shows the results for a dense ePUF (pattern #198). Pattern matching was conducted with distance ratio k=0.7, prominence threshold p=0.5, matching threshold λ=0.02, and symmetry checking. Number of feature matches are 90 and 154 for (a) and (b), respectively.
Figure 8
Figure 8
A collection of images captured with a cellphone camera at various orientations and lighting conditions to test the accuracy and robustness of the pattern matching algorithm. Pattern matching to the database images was conducted using k=0.7 (distance ratio), p=0.5 (prominence threshold), λ=0.02 (matching threshold), with symmetry checking. Label in the top left (correct/unconfident) indicates if the image accurately matches with the database image for that tablet. Unconfident means the matching prominence fell below the threshold. Label in the bottom left is the number of correct feature matches (to the best match in the database), over the total number of features detected by SIFT for that image. Mean brightness of the images range from 130 to 200.
Figure 9
Figure 9
Cross-validation results for 50 randomly-chosen images from the database. Left plot (a) shows the pattern matching results with a prominence threshold p=0 (i.e., no thresholding), and right plot (b) is with p=0.5. All pattern matching (a and b) was done with symmetry checking enabled. Both plots were computed with a matching threshold λ=0.02.
Figure 10
Figure 10
Effects of distance ratio k and symmetry checking on specificity of feature matching. Plot (a) shows the effect for a sample image from the database, checked against the database images (200 in total). Plot (b) shows the effect for a cellphone image of the same sample (ePUF pattern #94; same ePUF pattern as in (a)) checked against the database images. Colored lines represent different feature matching conditions (k-value and symmetry checking) as specified in the legends. The peaks in (a) are all the same value (as labeled), and each of the peaks in (b) are labeled with their respective color.
Figure 11
Figure 11
Heat maps showing the effect of distance ratio (in feature matching), and prominence threshold (for ePUF matching) for (a, c) the database images (checked against themselves), and (b, d) the cellphone images (checked against the database). Plots in the row (a, b) were computed without symmetry checking, and the bottom row (c, d) with symmetry checking. Colors represent different values of η (product of precision and recall), as represented in the color bar on the right. Pattern matching was conducted without symmetry checking. All results computed with a matching threshold λ=0.02.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Kovacs S, Hawes SE, Maley SN, Mosites E, Wong L, Stergachis A. Technologies for detecting falsified and substandard drugs in low and middle-income countries. PLoS ONE. 2014;9(3):e90601. doi: 10.1371/JOURNAL.PONE.0090601. - DOI - PMC - PubMed
    1. Bansal D, Malla S, Gudala K, Tiwari P. Anti-counterfeit technologies: A pharmaceutical industry perspective. Sci. Pharma. 2013;81:1–13. doi: 10.3797/scipharm.1202-03. - DOI - PMC - PubMed
    1. Kumar G. Pharmaceutical drug packaging and traceability: A comprehensive review. Univ. J. Pharm. Pharmacol. 2023;2(1):19–25. doi: 10.31586/ujpp.2023.769. - DOI
    1. Coustasse A, Arvidson C, Rutsohn P. Pharmaceutical counterfeiting and the RFID technology intervention. J. Hosp. Mark Public Relat. 2010;20(2):100–115. doi: 10.1080/15390942.2010.493369. - DOI - PubMed
    1. Saim Gunay U, Eric Petit B, Karamessini D, Gigmes D, Charles L, Ois Lutz J-F. Chemoselective synthesis of uniform sequence-coded polyurethanes and their use as molecular tags. Chem. 2016 doi: 10.1016/j.chempr.2016.06.006. - DOI