Implementation of volatile organic compound identification algorithms using

Implementation of volatile organic compound identification algorithms using colorimetric sensor array data Alexandra Stephens Mentored by Dr. Alan Samuels and Dr. Charles Davidson Introduction Results(Continued) Permethrin Signature The angle-between-vectors approach was the most successful, only misidentifying two chemicals out of the 34. The dot-product was the least successful, as most chemicals were identified as either Bleach, Permethrin, or Hoppes #9, chemicals with generally high magnitudes. The remaining two codes performed relatively well, each identifying 29 to 30 chemicals correctly. Conclusion 1. 5 RGB Value Colorimetric sensor arrays create a way to “see smells. ” These small “tickets” consist of 76 colored spots with different chemical compositions, such as metalloporphyrins and hydrogen bonding sites. When exposed to volatile organic compounds (VOCs) or other chemicals, the spots’ molecular structures foster various intermolecular reactions, ranging from Lewis donor/acceptor reactions to Brønsted acid/base reactions (Suslick, 2004). The result of these chemical changes is reflected in the change in color of the dots. The red, green, and blue (RGB) values of each spot are extracted through digital color imaging n times until the reaction is complete, creating an n by 228 (76 times three) matrix. Analysis of VOC ticket data is utilized in chemical identification; some methods of identification currently include computing the dot product of data sets, the k nearest-neighbor algorithm, and hierarchical cluster analysis. However, a permanent and accurate algorithm has yet to be established. The purpose of this investigation was to develop an approach that analyzes colorimetric sensor array data and correctly identifies at least 90% of chemicals. Methods and Materials (Continued) 1. 0 0. 5 0. 0 -0. 5 -1. 0 0 50 100 150 200 Band Number Graph 1: This is a graph of the change in the color values of each of the 76 spots on a colorimetric sensor array over time when exposed to the common household item, Permethrin. Results Methods and Materials Scaled RGB Value Scaled Permethrin Signature 1. 0 0. 5 0. 0 -0. 5 -1. 0 0 50 100 Band Number 150 200 Graph 2: This graph shows the signature of Permethrin after it is altered by the first program. Percent Correctly Identified Chemical Testing Results 100% 85, 29% 94, 12% 82, 35% References 60% 40% 20% 5, 88% 0% Dot Product One-Zero Comparison Z-Score The purpose of this study was to develop a program that analyzes colorimetric sensor array data and correctly identifies at least 90% of the 34 household chemicals given. The angle-between-vectors approach surpassed the 90% accuracy goal, and drastically improved upon the dot product approach. This is because it eliminates the potential to incorrectly identify a VOC due to high magnitudes of color change, found in chemicals such as Bleach. The one-zero method did surprisingly well, given the simplicity in the program. It performed slightly better than the z-score approach, which is more complex and uses more advanced statistics. The same issue arose in all of the identification programs: since some of these household substances were quite similar, for example, two different versions of OFF® insect repellent were tested, many of these substances identified as one another. This may be because the actual chemical compositions of these substances are so similar, the data varies only slightly, causing confusion in some or all of the identification algorithms. To advance this study, larger sample sizes should be used to ensure consistency of the programs. It is an important piece to the many applications of colorimetric sensor array data analysis. For example, lung cancer and other diseases can be identified through analysis of the breath of patients with colorimetric sensor arrays (Beukemann et. al. , 2012). They serve as a less-invasive, less-expensive, and potentially more accurate diagnostic tool. This situation can be life threatening, and an identification program with high accuracy (at least 90%) is necessary. Angle Between Vectors Graph 3: This graph shows the results of testing an old identification method, the dot product, and the three new methods with 34 different colorimetric sensor array VOC data sets. Beukemann, M. C. , Kemling, J. W. , Mazzone P. J. , Mekhail, T. , Na, J. , Sasidhar, M. , …Xu, Y. (2012). Exhaled breath analysis with a colorimetric sensor array for the identification and characterization of lung cancer. J Thorac Oncol, 7(1): 137– 142 doi: 10. 1097/JTO. 0 b 013 e 318233 d 80 f Suslick, K. S. (2004). An optoelectronic nose: “seeing” smells by means of colorimetric sensor arrays. MRS Bulletin. Retrieved from www. mrs. org/publications/bulletin