It is widely recognized that bloodstains dating may represent a crucial evidence in resolving many cases of forensic interest. Since the non-destructive techniques are most desirable, spectroscopies appear to be particularly attractive. Even though this topic has been investigated with limited success since the 1960s, most scientific strain focused on UV-Vis spectroscopy, while few studies were conducted in the IR spectral region, with only a handful of them concerning NIR spectroscopy. The final goal of the present work is to compare the performances of NIR spectroscopy for bloodstains dating with those of UV-Vis in the prospect of real-casework implementation. To reach this aim, capillary blood was collected from a healthy volunteer, deposited onto glass substrates, and stored under stable laboratory conditions. The aging of bloodstains was followed for 16 days measuring transflection NIR spectra and diffuse-transmittance UV-Vis spectra (6 bloodstains for each technique, analyzed in triplicate). The resulting data were independently analyzed by several chemometric techniques. Preprocessing of spectral profiles was conducted first with classical row pre-processing methods, such as Savitzky-Golay smoothing and standard normal variate (SNV) transform. After conducting an exploratory analysis by principal component analysis (PCA), systematic differences between samples due to factors unrelated to aging (such as blood deposition thickness) were detected as predominant. Therefore, class centering was applied to remove this unwanted effect yielding the resulting data to be focused only on time trends. Subsequently, a partial least square (PLS) regression model was computed to evaluate the effectiveness of applied spectral methods for estimating time elapsed since blood trace deposition. The two techniques gave comparable root means square errors in prediction (RMSEP) equal to about one day. This time uncertainty appears to be lower than any other so far reported in the literature for NIR spectroscopy. In view of the creation of a multi-instrumental platform for bloodstains dating, data fusion strategies were also tested to estimate the advantages and disadvantages of low-level, mid-level, and high-level approaches. The present results together with the regression models obtained from them looks promising, even taking into account the limit of the lack of variability of the blood donor. NIR spectroscopy is therefore seemingly worthy of increased scientific attention in the forensic field for bloodstains dating.
Bloodstains dating by means of NIR and UV-VIS spectroscopy - a critical comparison
Sara Gariglio;Cristina Malegori;Paolo Oliveri
2023-01-01
Abstract
It is widely recognized that bloodstains dating may represent a crucial evidence in resolving many cases of forensic interest. Since the non-destructive techniques are most desirable, spectroscopies appear to be particularly attractive. Even though this topic has been investigated with limited success since the 1960s, most scientific strain focused on UV-Vis spectroscopy, while few studies were conducted in the IR spectral region, with only a handful of them concerning NIR spectroscopy. The final goal of the present work is to compare the performances of NIR spectroscopy for bloodstains dating with those of UV-Vis in the prospect of real-casework implementation. To reach this aim, capillary blood was collected from a healthy volunteer, deposited onto glass substrates, and stored under stable laboratory conditions. The aging of bloodstains was followed for 16 days measuring transflection NIR spectra and diffuse-transmittance UV-Vis spectra (6 bloodstains for each technique, analyzed in triplicate). The resulting data were independently analyzed by several chemometric techniques. Preprocessing of spectral profiles was conducted first with classical row pre-processing methods, such as Savitzky-Golay smoothing and standard normal variate (SNV) transform. After conducting an exploratory analysis by principal component analysis (PCA), systematic differences between samples due to factors unrelated to aging (such as blood deposition thickness) were detected as predominant. Therefore, class centering was applied to remove this unwanted effect yielding the resulting data to be focused only on time trends. Subsequently, a partial least square (PLS) regression model was computed to evaluate the effectiveness of applied spectral methods for estimating time elapsed since blood trace deposition. The two techniques gave comparable root means square errors in prediction (RMSEP) equal to about one day. This time uncertainty appears to be lower than any other so far reported in the literature for NIR spectroscopy. In view of the creation of a multi-instrumental platform for bloodstains dating, data fusion strategies were also tested to estimate the advantages and disadvantages of low-level, mid-level, and high-level approaches. The present results together with the regression models obtained from them looks promising, even taking into account the limit of the lack of variability of the blood donor. NIR spectroscopy is therefore seemingly worthy of increased scientific attention in the forensic field for bloodstains dating.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
