Advanced computer Vision techniques for drug abuse detection

Lateral flow tests, used to rapidly detect various diseases, as HIV, or specific physiological conditions, as drug abuse, through blood, saliva, or urine samples, are becoming a powerful and cost-effective diagnostic tool. One major factor affecting the test result is the subjectivity of the operator's reading, which relies on both the interpretation of the results and the assessment of sample compliance. To overcome this issue, Computer Vision (CV) provides tools to mitigate the subjectivity of the results. Indeed, through sophisticated CV algorithms, it is possible to calibrate and normalize the result interpretation, taking into account individual variations [1] and environmental influences. In this talk, we present an automated lateral flow test reader for drug abuse detection, enabling both operator-independent interpretation of results and objective validation of sample compliance through CV techniques. One of the main challenges addressed in this study is to tackle the issue of non-uniform lighting in the analysis scene, while at the same time dealing with the variability in the positioning of the regions of interest. We propose an innovative method for objectively detecting the presence or absence of illicit substances, establishing a threshold for positivity and assessing the suitability of the analyzed sample, regardless of the limitations and subjectivity associated with the operator. A combination of filtering, image enhancement, and segmentation techniques were employed to extract relevant information. Additionally, color balancing and clustering methods were used to investigate the colors of sample suitability indicators. The results demonstrate the effectiveness of the proposed method in improving objectivity in rapid lateral flow test results.