• Users Online: 113
  • Home
  • Print this page
  • Email this page
Home About us Editorial board Ahead of print Current issue Search Archives Submit article Instructions Subscribe Contacts Login 


 
 Table of Contents  
ORIGINAL ARTICLE
Year : 2017  |  Volume : 3  |  Issue : 2  |  Page : 63-67

Discrimination of handlebar grip samples by fourier transform infrared microspectroscopy analysis and statistics


Collaborative Innovation Center of Judicial Civilization, Key Laboratory of Evidence Science, CUPL, Ministry of Education, China

Date of Web Publication30-Jun-2017

Correspondence Address:
Yuanfeng Wang
Collaborative Innovation Center of Judicial Civilization, Key Laboratory of Evidence Science, CUPL, Ministry of Education
China
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/jfsm.jfsm_54_17

Rights and Permissions
  Abstract 


In this paper, the authors presented a study on the discrimination of handlebar grip samples, to provide effective forensic science service for hit and run traffic cases. 50 bicycle handlebar grip samples, 49 electric bike handlebar grip samples, and 96 motorcycle handlebar grip samples have been randomly collected by the local police in Beijing (China). Fourier transform infrared microspectroscopy (FTIR) was utilized as analytical technology. Then, target absorption selection, data pretreatment, and discrimination of linked samples and unlinked samples were chosen as three steps to improve the discrimination of FTIR spectrums collected from different handlebar grip samples. Principal component analysis and receiver operating characteristic curve were utilized to evaluate different data selection methods and different data pretreatment methods, respectively. It is possible to explore the evidential value of handlebar grip residue evidence through instrumental analysis and statistical treatments. It will provide a universal discrimination method for other forensic science samples as well.

Keywords: Evidential value, Fourier transform infrared microscpectroscopy, handlebar grip sample, principal component analysis, receiver operating characteristic


How to cite this article:
Lin Z, Li B, Du R, Wei Z, Wang Y. Discrimination of handlebar grip samples by fourier transform infrared microspectroscopy analysis and statistics. J Forensic Sci Med 2017;3:63-7

How to cite this URL:
Lin Z, Li B, Du R, Wei Z, Wang Y. Discrimination of handlebar grip samples by fourier transform infrared microspectroscopy analysis and statistics. J Forensic Sci Med [serial online] 2017 [cited 2017 Aug 22];3:63-7. Available from: http://www.jfsmonline.com/text.asp?2017/3/2/63/209291




  Introduction Top


In many countries, bicycles, the electric bikes, and the motorcycles play important roles in the modern transport. The three common means of transport in China are practical and economical. However, more and more bicycles/electric bikes/motorcycles have been involved in traffic cases. Meanwhile, serious injuries have frequently happened to the rider because most of them did not wear helmet. Sometimes, there was indeed collision between the perpetrator and accident victim or their vehicles. In this case, the handlebar grip residue from the bicycles/electric bikes/motorcycles always transferred to the surface of suspect vehicle and leave black traces on the other object during the collision. It is because that the handlebar grips stretch out quite much and very easily to be touched by another object. Whereas, there was sometime no collision between the victim and suspect vehicles during the accident. Victim fell down because he/she changed the direction suddenly during the accident to avoid collision. In this case, the grip material contamination from daily life might confuse the fact finder. Different facts will lead to different judgements in the courtroom. Thus, it is necessary to set up a scientific method for tracing the fact in hit and run cases. Especially, we need to be confident to distinguish the situation with slight collision from the situation without collision.

Fourier transform infrared microspectroscopy (FTIR) was frequently utilized in the analysis of micro trace evidence.[1] FTIR spectrum could reflect the characteristic of chemical structure of the specimen, due to that (i) the molecules absorb resonant frequencies that match the vibrational frequency and (ii) the absorption intensity is affected by the shape of the molecular potential energy surfaces, the masses of the atoms, and the associated vibronic coupling. Therefore, FTIR became a worldwide analytical tool for forensic science purpose. The evidential information explored from the fingerprint region of FTIR spectrum could remarkably improve the discrimination of different chemical components.

However, what we encountered more frequently in the practice of micro trace evidence examination is that the unknown sample and the compared one always share the similar organic components,[2] which will lead to similar FTIR spectrums. The difference between them was minor and easily ignored. It was resulted from the mass production in modern society. Since FTIR spectrum is not sensitive enough to reflect the minor components (<5%) in the mixture, we should not only rely on FTIR method but also other exploring tools (such as statistics),[3],[4] so as to search for the evidential information hidden behind the common appearance of forensic science samples and differentiate the similar ones effectively.


  Materials and Methods Top


Specimen preparation

Since bicycles, electric bikes, and motorcycles are frequently involved in hit-and-run traffice cases; we randomly collected handlebar grip samples from 50 bicycles, 49 electric bikes, and 96 motorcycles from the local police (Beijing, China). All these bicycles, electric bikes, and motorcycles were involved traffic accident cases before 2012. Therefore, we could suppose that none of the 195 handlebar grip samples was from the same source with the others.

Fourier transform infrared microspectroscopy microspectroscopy

A Nicolet Spectrum 6700 FTIR Spectrometer with a Nicolet Continuμm ™ Infrared Microscope and a narrow band mercury cadmium telluride detector was used to collect the FTIR spectra under the reflection mode. The instrument was operated at a 4/cm resolution over a spectral range of 4000–650/cm.

FTIR spectrum of the 195 handlebar grip samples was collected under the aforementioned experimental conditions. For each handlebar grip sample, we prepared three FTIR samples and conducted five replicate analysis for each of the three samples. Therefore, each handlebar grip will finally obtain 15 FTIR spectrums. The results of traditional FTIR analysis indicated that (i) 0.5% of them were made of polyamide, (ii) 0.5% of them were made of polyethylene-polypropylene copolymer, (iii) 0.5% of them were made of polyethylene, (iv) 7.6% of them were made of polyester, and (v) 90% of them were made of the mixture of calcium carbonate and polymer.

Optimization of the discrimination

To optimize the discriminative power of FTIR method when applied to forensic examination of handlebar grip samples, we followed the three steps as shown in [Figure 1]. Four data selection methods, three data pretreatment methods and eight discrimination methods were utilized to achieve a better result. According to the literatures,[5],[6] data selection and data pretreatment will help to filter out useless data, reduced the influence of less important data and focus on more important ones. Especially when we have two sets of date for comparison and both of them are multidimensional data, this optimization procedure will remarkably increase the discriminating power of the specific instrumental analytical method, such as FTIR spectrum. The numerical data obtained by FTIR analysis was statistically analyzed with R ® (Version 3.2.5).
Figure 1: Optimization procedure of the discrimination

Click here to view



  Results and Discussion Top


Evidential information based on

When a bicycle, electric bike or motorcycle was involved in a hit-and-run traffic accident, we always collected the residues deposited on the suspect vehicle's surface as unknown sample and collected the handlebar grip material as comparing sample. According to the data shown in [Table 1], some types of the handlebar grip materials have very high rarity. For example, for polyamide, it only appeared once in a bicycle handlebar grip sample. Therefore, the data shown in [Table 1] provide us with highly important information if what we met in casework was polyamide, polyester, polypropylene, or polyethylene-polypropylene copolymer. However, as aforementioned in the part 2.2, if both of the unknown sample and the comparing sample were the mixture of calcium carbonate and polymer, it is difficult to tell if they are from the same source or not, according to their FTIR spectrum comparison. We need to find out a way to go deeper into the data, investigate the sample more carefully, and provide more evidential information from it. Therefore, it is necessary to improve the discriminating power of FTIR analysis when examining micro trace evidence with similar components.
Table 1: Classification of 195 handlebar grip samples by traditional Fourier transform infrared microspectroscopy analysis

Click here to view


Selection of target absorptions

In our experiment, FTIR spectrum covered a broad wavelength range from 4000 cm to 650 cm with a step of 4/cm. Therefore, for each sample, the data set was composed of 936 data points and became a multidimensional data. To reveal the minor difference between handlebar grip samples through FTIR spectrum and remove interference from the wavelength range with less evidential value, we chose four types of target absorptions to evaluate their discriminating power through principal component analysis (PCA). As shown in [Table 2], the four types of target absorptions were 4000–650/cm (TS-1), two main interesting areas of 3100–2700/cm and 1900–650/cm (TS-2), ten smaller interesting areas (TS-3), and 16 target peaks (TS-4).
Table 2: Four types of target absorptions selection

Click here to view


We chose two samples from the sample group. As described in 2.2, each sample was collected 15 spectrums. A PCA was carried out on the data from these two samples to highlight potential discriminative power of different data sets. Sample-1 and sample-2 were presented with red color and green color, respectively. As shown in [Figure 2], distinct group of samples could be observed under the four types of data sets. However, the distances between two sample groups were different among them. Generally speaking, TS-3 and TS-4 presented relatively better results than TS-1 and TS-2 because dots for sample-1 (red color) are more grouped, and the distance between two samples is much bigger. It indicated that TS-3 and TS-4 properly focused on the interesting area which functional groups were sensitive to. Therefore, we temporarily kept these two of target absorptions for next steps.
Figure 2: Principal component analysis of sample-1 (red color) and sample-2 (green color) under four different data sets. a: TS-1; b: TS-2; c: TS-3; d: TS-4

Click here to view


Statistical treatments

Normalization, square root and fourth square root were applied on the selected target absorptions under TS-3 and TS-4, respectively, to reduce the influence of larger peaks through pretreatment. Meanwhile, discriminant analysis was performed to investigate the possibility of discriminating different handlebar grip samples. Three correlation methods (Pearson, Kendall and Spearman) and five distance methods (Euclidean, Maximum, Manhattan, Canberra and Minkowski) generally used in the field of forensic science were retained to process the pretreated data. There were altogether 48 combinations as shown in [Table 3].
Table 3: Values of area under the receiver operating characteristic curves

Click here to view


One hundred and ninety-five handlebar grip samples were divided into two groups, including twenty handlebar samples assigned to linked samples and 175 handlebar samples assigned to unlinked samples. As for linked samples, 15 FTIR spectrums from replicate analysis of each sample were compared with each other. As for unlinked samples, comparisons were conducted between FTIR spectrums from different samples. To evaluate the separation between the distributions of linked samples and unlinked samples, receiver operating characteristic (ROC) curves were used. The area under the ROC curve was calculated on the basis of each combination of a specific selected absorption targets, pretreatment, and correlation/distance function [Table 3].

Since the ROC curves were built by plotting the true positive rate (sensitivity) as a function of the false positive rate (1-specificity) at each correlation value, the area under the ROC curve (theoretically between 0.5 and 1) allows the quantification of overlapping degree between the compared distributions and evaluation of each combination of statistical treatments. The value of 0.5 means distribution completely overlapped; while, the value of 1 stands for distribution entirely separated. Generally speaking, the data shown in [Table 3] indicated that (i) the separation based on TS-4 yielded better results than TS-3, (ii) the five distance functions applied in this study yielded better identical results than the three correlation functions applied in this study, and (iii) the difference between pretreatments was not obvious. The best combination of statistical treatments was the combination of data set TS-4, fourth square root and Canberra distance yielding a ROC area of 0.9994.

Separation of linked and unlinked samples

The distribution of linked samples (n = 20) and unlinked samples (n = 175) was conducted through the best statistical treatment as described in 3.3. As shown in [Figure 3], histogram presented Canberra distances between samples based on data set TS-4 with pretreatment of the fourth square root. For the two populations, the linked ones grouped closely to the left areas [Figure 3], left], while the unlinked ones widely distributed in the other areas [Figure 3], right]. It indicated that the linkage between samples can ben searched by the optimized statistical treatment introduced in this paper. It not only filters out the unnecessary noise in FTIR spectrum but also keeps the signals with evidential information efficiently and enlarges them effectively.
Figure 3: Histogram of distributions of Canberra distances calculated between linked and unlinked samples, where data were chosen by TS-4 and pretreated by fourth square root

Click here to view



  Conclusion Top


This study aimed to explore the potential evidential information hidden inside the FTIR spectrum of handlebar grip samples, especially the similar ones. According to the traditional FTIR analysis, more than 90% of the handlebar grip samples were made of the mixture of calcium carbonate and polymer. It not only indicated the difficulties that forensic examination of handlebar grip samples will meet but also reflected the problems that many forensic scientists will encountered in the other subfield of forensic science.

Statistical tools were utilized to improve the advanced analysis of FTIR spectrum. The choice of target selected peaks, data pretreatment methodologies, and discriminant analysis was decided after a systematic research. Among these three factors, different target selected peaks and discriminant analysis had great influence on the final distributions of samples, whereas different pretreatment methodologies presented much less distinction. The results indicated that combination of TS-4 data set, pretreatment of fourth square root, and Canberra distance presented a better discriminating power than other options.

The source-level hypotheses we commonly encountered in a case circumstance are “the compared samples were from the same source (Hp)/or not (Hd).”[7] However, on the one hand, differences exist among different parts of the same object; on the other hand, similarities appear between different samples. The optimized statistical treatment introduced an objective approach to measure the distance between multi-dimension chemical profiles. The exploration of more characteristic, higher specificity, and less uncertainty will be fully implemented with these forensic science tools.

Acknowledgment

This work was financially supported by Beijing Nova Programme (Grant Number: Z1511000003150123), China, and Key Program of National Social Science Fund (Grant Number: 16AYY015).

Financial support and sponsorship

This work was financially supported by Beijing Nova Programme (Grant Number: Z1511000003150123), China, and Key Program of National Social Science Fund (Grant Number: 16AYY015).

Conflicts of interest

There are no conflicts of interest.



 
  References Top

1.
Wang Y, Li B. Determination of the sequence of intersecting lines from laser toner and seal ink by Fourier transform infrared microspectroscopy and scanning electron microscope/energy dispersive X-ray mapping. Sci Justice 2012;52:112-8.  Back to cited text no. 1
    
2.
Yuanfeng WA, Yingchao YU, Guiling WU. Study on the application of statistics to scientific evidence report. Evid Sci 2016;24:506-13.  Back to cited text no. 2
    
3.
Robertson B, Vignaux GA. Interpreting Evidence: Evaluating Forensic Science in the Courtroom. 1st ed. UK: Wiley; 1995.  Back to cited text no. 3
    
4.
Taroni F, Aitken C, Garbolino P, Biedermann A. Bayesian Networks and Probabilistic Inference in Forensic Science Statistics in Practice. UK: Wiley; 2006.  Back to cited text no. 4
    
5.
Weyermann C, Marquis R, Delaporte C, Esseiva P, Lock E, Aalberg L, et al. Drug intelligence based on MDMA tablets data I. Organic impurities profiling. Forensic Sci Int 2008;177:11-6.  Back to cited text no. 5
    
6.
Marquis R, Weyermann C, Delaporte C, Esseiva P, Aalberg L, Besacier F, et al. Drug intelligence based on MDMA tablets data: 2. Physical characteristics profiling. Forensic Sci Int 2008;178:34-9.  Back to cited text no. 6
    
7.
European Network of Forensic Science Institutes. ENFSI Guideline for Evaluative Reporting in Forensic Science – Strengthening the Evaluation of Forensic Results Across Europe. Belgium: European Network of Forensic Science Institutes; 2015. p. 4-29.  Back to cited text no. 7
    


    Figures

  [Figure 1], [Figure 2], [Figure 3]
 
 
    Tables

  [Table 1], [Table 2], [Table 3]



 

Top
 
 
  Search
 
Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

 
  In this article
Abstract
Introduction
Materials and Me...
Results and Disc...
Conclusion
References
Article Figures
Article Tables

 Article Access Statistics
    Viewed251    
    Printed2    
    Emailed0    
    PDF Downloaded23    
    Comments [Add]    

Recommend this journal


[TAG2]
[TAG3]
[TAG4]