Colorimetric parameters for bloodstain characterization by smartphone

Abstract

This smartphone colorimetry investigated RGB (red, green and blue), HSV (hue, saturation and value of brightness) and CMYK (cyan, magenta, yellow and key) color parameters for characterizing bloodstains. Blood samples, separately deposited on cotton cloths and tissue papers, were analyzed by the ‘Colorimeter X’ application on an iPhone XR after 1, 5, 7, 9 and 11 days. The reductions in RGB and V values recorded by the smartphone application are consistent with darker bloodstain over time. Bloodstain colors are sensitive to the substrate and person-to-person variation, but a similar correlation with time could be established by a linear least squares fitting. With the highest coefficient of determination averaged from four samples, the V value is the most effective parameter for describing bloodstain ageing. For other color spaces, the R and K almost match the V value in measuring the bloodstains. Because these three values may be susceptible to the deviation in the measurement, the B value is suggested as an additional parameter in the bloodstain characterization. This small-scale study demonstrates the potential of colorimetric parameters in an on-site analysis by a cost-effective smartphone.

Keywords:

• Bloodstain
• color model
• digital image colorimetry
• image analysis
• smartphone

1. Introduction

Digital image colorimetry is an efficient quantitative analysis of colors from images. Modern digital devices mostly form the digital images according to the RGB (red, green and blue) color space, with each component ranging from 0 to 255. However, other colorimetric systems are more suitable in some applications (Apyari, Gorbunova, Isachenko, Dmitrienko, & Zolotov, 2017; Marrone, La Russa, Montesanto, Lagani, La Russa, & Pellegrino, 2021; Shin et al., 2017; Thanakiatkrai, Yaodam, & Kitpipit, 2013). Therefore, the RGB color space is often converted into the HSV (hue, saturation and value of brightness), HSL (hue, saturation and lightness), CMYK (cyan, magenta, yellow and key) and CIELAB/CIELCh color spaces (Gonzalez & Woods, 2018; Pham et al., 2007).

With the proliferation of digital cameras, smartphone cameras, and other devices that can record images, a method for assessing the color characteristics of an image has been implemented in scientific works. Apyari et al. (2017) summarized the use of color-recording devices and colorimetric systems for analyzing water, tea, blood, and other chemical substances. Many colorimetric sensor arrays have been developed to capture the color components of samples. Ma, Li, Ma, and Wang (2018) reviewed the types of optical colorimetric sensors applied in the chemical and biological analysis of organic molecules, DNA, amino acids, proteins, etc. Digital image colorimetry with built-in smartphone cameras is an accessible, rapid and cost-effective method to measure the colors of digital images (Fan, Li, Guo, Xie, & Zhang, 2021). The smartphones enable the on-site quantitative analysis of chemical substances (Danyliuk, Tatarchuk, Kannan, & Shyichuk, 2021; Zheng et al., 2022), wastewater (Santiago & Sevilla, 2022; Thio, Bae, & Park, 2022), foods (Kumar, Aulakh, Gill, & Sharma, 2022; Nisoa, Wattanasit, Tamman, Sirisathitkul, & Sirisathitkul, 2021; Patel et al., 2022; Wu, Zhang, Kuang, Fang, & Song, 2022; Zhang et al., 2022), fabrics (Ramirez & Santamarina, 2022; Sirisathitkul & Kaewareelap, 2021), skin (Cugmas, Štruc, Kovče, Lužar, & Olivry, 2022; Cugmas & Štruc, 2020), ancient potteries (Sirisathitkul, Ekmataruekul, Sirisathitkul, & Noonsuk, 2020) and forensic evidence (Choi, Shin, Hyun, Song, & Jung, 2019; Marrone et al., 2021; Shin et al., 2017; Thanakiatkrai et al., 2013).

Bloodstains often left at crime scenes are vital pieces of forensic evidence. After the blood leaves the human body, the change in physical characteristics over time is remarkable. The blood coagulates in its pool, and becomes a dry bloodstain on a substrate after the gelation and desiccation from the edge to the center. Also, the crack propagates in the bloodstain pattern over time. The blood is initially bright red once oxygenated in the air. It gradually turns into a dark brown stain due to the change from oxy-hemoglobin to methemoglobin and, eventually, hemichrome. These derivatives of hemoglobin, the majority of blood’s dry content, absorb and reflect visible lights of different wavelengths (Choi et al., 2019; Thanakiatkrai et al., 2013). The period from their formation to the analysis indicates the time of incidence, linking to suspects in the violent crime. The postmortem interval of how long it has been since the victim’s death can be determined by the time since the deposition of blood (Cavalcanti & Silva, 2019; Shin et al., 2017). Although the bloodstain age up to about 3 weeks is highly relevant to forensic investigations (Bergmann, Leberecht, & Labudde, 2021), it is intricate to accurately measure in the real case due to several factors involved. Laboratory techniques capable of estimating the bloodstain age are Raman Scattering (Sharma & Kumar, 2018; Weber & Lednev, 2020), Fourier Transform Infrared (FTIR) spectroscopy (Sharma & Kumar, 2018; Zadora & Menżyk, 2018), Ultraviolet/Visible (UV/Vis) spectroscopy (Bergmann, Heinke, & Labudde, 2017; Zadora & Menżyk, 2018), visible reflection spectroscopy Sun et al., 2017), fluorescence spectroscopy (Weber, Wójtowicz, & Lednev, 2021), atomic force microscopy (Cavalcanti & Silva, 2019), liquid chromatography-tandem mass spectrometry (Kim et al., 2022), real-time polymerase chain reaction (PCR) (Manasatienkij & Nimnual, 2021), analysis of Ethylenediaminetetraacetic acid (Bergmann et al., 2021), RNA (Heneghan, Fu, Pritchard, Payton, & Allen, 2021) and MicroRNAs (Fang et al., 2020). However, each method has its limitation due to many factors influencing the measurements (Zadora & Menżyk, 2018). Furthermore, these methods cannot be carried out on-site due to the need for specialized laboratory instruments, time-consuming analysis, and specialists.

For colorimetric studies of bloodstains, Marrone et al. (2021) recorded CIELAB/CIELCh colors over time using a spectrophotometer. With color changes from bright red to dark brown, models for predicting the time since deposition of blood were constructed from the correlation of color values with the age of bloodstains. Thanakiatkrai et al. (2013) used a computer to analyze images captured by a smartphone camera in a dedicated light box and then traced the time-dependent colorimetric parameters in RGB, HSV, HSL and CMYK color spaces. The bloodstain age was then estimated using the M value based on its linear decrease with time on the logarithm scale. The effects of temperature, humidity, light exposure, anticoagulant, and type of substrate were investigated because the rate of bloodstain color change is highly dependent on these factors (Thanakiatkrai et al., 2013). ‘Smart Forensic Phone’, an Android-based smartphone application for estimating the age of bloodstains, was developed by a research group at Yonsei University (Shin et al., 2017). The time since the deposition of blood up to 42 h was rapidly determined from the decreases in the V value, corresponding to the brightness in the HSV color space. This system is highly promising for an on-site analysis at the crime scene. The research group at Yonsei University subsequently developed the additional pattern recognition and classification of blood pool and crack in the smartphone image to estimate the bloodstain age, whereas the V value was still used in the colorimetric analysis (Choi et al., 2019).

This study compares the color variations of bloodstains to determine the essential parameters for quantitative description of the blood color. The focus on the period for up to 11 days is aimed to explore a linear correlation after the rapid color changes phase within the initial 42 h (Shin et al., 2017). Three different color spaces were investigated in this smartphone colorimetry because the essential parameters used in the literature are different (Choi et al., 2019; Marrone et al., 2021; Shin et al., 2017; Thanakiatkrai et al., 2013). Based on the dependence of hemoglobin level on the blood type reported in some published articles (Kumar & Kumar Singh, 2020; Ramalingam & Raghavan, 2020), samples from different blood types are aimed to provide different starting colors in this study. Cloth and paper were selected for blood deposition because the pattern on these absorbent substrates cannot be analyzed, and the characterization must rely on color quantification. Both substrates are white to minimize their interference with the colorimetry (Thanakiatkrai et al., 2013).

2. Materials and methods

Bloodstains covering all four blood types, referred to as samples A, B, AB and O, were collected from four healthy female volunteers. The donors from a purposive sampling have an average age of 21.19 ± 0.31 years. The levels of hemoglobin are 13.5, 14.0, 12.7 and 13.0 g/dL for the donors of samples A, B, AB and O, respectively. The blood from the fingertip was separately deposited on white sheets of cotton cloth and tissue paper, as exemplified in Figure 1a and b. The samples were kept at room temperature in sealed plastic bags without exposure to light and then used in the measurement after the deposition for 1, 5, 7, 9 and 11 days. For digital image colorimetry, the ‘Colorimeter X’ application developed by Dmitry Svishchov was downloaded from the App store and installed on an Apple smartphone (iPhone XR). Like other mobile colorimetric applications, the ‘Colorimeter X’ displays color components of an image acquired by a smartphone according to the RGB, CYMK, CIELAB, HSL and HSV color spaces. It can also facilitate the color mixing and comparison. To assess the validity of smartphone colorimetry, the repeated measurements on standard color cards (Pasco Color Mixer Accessory Kit OS-8495; red Pantone 032U, pink Pantone 212U, and orange Pantone 137U) were compared to the values averaged from five readings by a ‘Pinetools’ program on a personal computer.

Figure 1. Photographs of bloodstains on (a) cotton cloth compared to (b) tissue paper, and (c) colorimetry setup.

The bloodstains on cotton cloths and tissue papers in this study were photographed by the smartphone on an adjustable stand, 13 cm above the samples under the fluorescent light. Although Shin et al. (2017) suggest that the smartphone colorimetry is not highly sensitive to the angle measurement, all samples were directly aligned with the smartphone on the vertical axis as shown in Figure 1c. Images in this study were taken without additional setting on the smartphone camera and processing by the ‘Colorimeter X’ application. Colors were exclusively measured at the center of each bloodstain. The colorimetric parameters in each color space were extracted from each image and plotted as a function of the time since deposition up to 11 days. Data points were fitted to the linear least squares regression line.

3. Results and discussion

Color values are sensitive to the illumination and camera setting for capturing and measuring digital images. The same lighting, camera and application were therefore used throughout this study. The RGB and HSV values were only used relatively to establish temporal variations. Nevertheless, the accuracy of smartphone colorimetry can be assessed by comparing to the measurement by a spectrophotometer as exemplified in the literature (Sirisathitkul et al., 2020). Digital photographic colors are also subjected to the white balance and other image normalizations by the camera and colorimetric application. In this study, the auto white balance setting allows the smartphone camera to determine an appropriate adjustment of the color temperature for objects on the white background. This can be beneficial for further implementing the bloodstain colorimetry under different types of lighting sources. The comprehensive normalization algorithms may significantly deviate the values from original colors. To examine the effect of color normalization by the ‘Colorimeter X’ application, RGB and HSV values read by the smartphone and ‘Pinetools’ by importing the same images into a personal computer are compared in Table 1. Overall values are comparable for both methods. The exception is 'Pinetools’ giving much lower S values. Moreover, B values are greatly different in the case of the red and orange cards. Small standard deviations of colorimetric values in Table 1 also confirm the repeatability of the smartphone colorimetry.

Table 1. Comparison of colorimetric values measured from standard color cards using Colorimeter X on smartphone and Pinetools on personal computer.

	Colorimetric parameter
Measurement method	R	G	B	H	S	V
	Color card: Red (Pantone 032 U)
Colorimeter X	168.52 ± 2.26	24.00 ± 2.77	9.64 ± 2.63	5.44 ± 1.16	94.24 ± 1.59	66.12 ± 0.88
Pinetools	154.20 ± 1.61	38.28 ± 2.15	24.6 ± 1.94	7.03 ± 0.75	84.12 ± 1.24	60.48 ± 0.65
	Color card: Pink (Pantone 212 U)
Colorimeter X	176.40 ± 1.58	63.68 ± 1.86	88.96 ± 1.43	346.36 ± 0.49	87.00 ± 0.96	69.16 ± 0.69
Pinetools	163.76 ± 2.09	71.68 ± 2.34	90.24 ± 2.13	347.92 ± 0.21	56.12 ± 1.17	64.2 ± 0.71
	Color card: Orange (Pantone 137 U)
Colorimeter X	196.16 ± 0.85	147.04 ± 1.65	0.2	45.16 ± 0.37	99.80 ± 0.41	76.80 ± 0.50
Pinetools	191.44 ± 1.26	151.32 ± 1.89	46.12 ± 1.79	43.50 ± 0.40	75.88 ± 0.73	75.24 ± 0.44

Figures 2a–5a show the reduction in the brightness of every blood sample with the time up to 11 days. The changes in bloodstain colors from deep red to brown are consistent with degrading hemoglobin derivatives, eventually into hemichrome (Choi et al., 2019; Thanakiatkrai et al., 2013). Because only one sample was collected for each blood type just to cover the different antigen presence and absence on the red blood cell membrane, the person-to-person variation is not intended to link to the blood type. More samples in future research are necessary to establish a correlation, if any, of each colorimetric parameter with the blood type. The person-to-person variation in Figures 2a–5a is consistent with the level of hemoglobin ranging from 12.7 to 14.0 g/dL. When red blood cells are separated by centrifugation, the remaining platelets, white blood cells and plasma are pale yellow. After drying, the color is not significantly changed. The color variation of bloodstains in this study is therefore attributed to the changes in red blood cells, similar to the literature (Choi et al., 2019; Marrone et al., 2021; Shin et al., 2017; Thanakiatkrai et al., 2013).

Figure 2. Changes of bloodstains from sample type A after 1–11 days (t), (a) photographic color, (b) R value plotted against ln t, (c) R value, (d) G value and (e) B value plotted against t.

Figure 3. Changes of bloodstains from sample type B after 1–11 days (t), (a) photographic color, (b) R value plotted against ln t, (c) R value, (d) G value and (e) B value plotted against t.

Figure 4. Changes of bloodstains from sample type AB after 1–11 days (t), (a) photographic color, (b) R value plotted against ln t, (c) R value, (d) G value and (e) B value plotted against t.

Figure 5. Changes of bloodstains from sample type O after 1–11 days (t), (a) photographic color, (b) R value plotted against ln t, (c) R value, (d) G value and (e) B value plotted against t.

The difference between bloodstains on cloth and paper is significant due to dissimilar texture and brightness of the substrates. Apart from sample B with the highest level of hemoglobin, the blood samples deposited on tissue paper appear darker. The temporal changes in blood color (Shin et al., 2017; Thanakiatkrai et al., 2013) and morphology of red blood cells (Cavalcanti & Silva, 2019) have been studied on different substrates, including metal, glass, wood, wallpaper, writing paper and cloth. For non-absorbent substrates, patterns of the blood pool and crack formation could also be correlated to the time since deposition, as demonstrated in Choi et al. (2019). Beside the substrate color, the effect of substrate on the blood colorimetry stems from the distribution in blood deposition and the electrical charge interaction at the interface (Cavalcanti & Silva, 2019). Any non-absorbent substrate such as ceramic tile results in a thicker layer than those on paper and cloth in this study.

The variations with time (t) and substrate are consistent with the RGB colors collected by the ‘Colorimeter X’ application. The blood color is characterized by the highest R value, whereas the G and B values are comparable. Since the color parameters in the literature decay exponentially in the initial period (Shin et al., 2017; Thanakiatkrai et al., 2013), the R value is firstly plotted against ln t in Figures 2b–5b. For sample A, the coefficient of determination (R²) from the least squares fitting is as high as 0.9822 on the cloth and 0.9902 on the paper. However, the plots of R value against ln t in other samples are not linearly fitted, as shown by the R² in Table 1. The plot of R value against time in Figures 2c–5c is considered a better fit for linear correlation, yielding a higher average R² for both substrates.

The G and B values, plotted with the time from 1 to 11 days in Figures 2d–5d, and Figures 2e–5e, also tend to decrease linearly. The R² averaged from the linear least squares fitting yields of G value from four samples and listed in Table 1, is 0.7665 on cloths and 0.9058 on papers. Notably, the trend in the G value resembles those of the R. In the case of sample O on cloth, both R and G values substantially deviate from a linear trend against time, as shown by Figure 5c and d and the R² in Table 1. In such a case, the B value in Figure 5e can still provide a linear correlation with a large negative slope. The R² averaged from four samples is 0.8798 on cloths and 0.8970 on papers. For this reason, the B value emerges as another essential parameter in blood characterizations.

The ‘Colorimeter X’ application also records color parameters in the HSV color space. Judging from the R² in Table 1, the H and S values are less effective than RGB in describing blood changes. The H value can be linearly fitted only in the case of sample A on cloth but fluctuates over time in other samples. Thanakiatkrai et al. (2013) also reported a low R² from the plot of H value against log t. By contrast, a decline in V values corresponding to an older blood’s darker tone is explicit. The V value from the bloodstain on tissue paper is less than that measured on the cloths, except for sample B. The plots of V value against time from 1 to 11 days in Figure 6, yield the average R² of 0.8254 on cloths and 0.9602 on papers. The efficiency of the V value in characterizing bloodstains is consistent with the implementations by Shin et al. (2017) and Choi et al. (2019). However, the V value is sometimes susceptible to the measurement deviation, and the R² drops below 0.5 for sample O on cloth. This finding is similar to the R value, highlighting the importance of an additional B value in the blood characterization.

Figure 6. The variation of V value with the time since deposition of blood up to 11 days (t) from sample type, (a) A, (b) B, (c) AB, and (d) O.

To investigate other parameters for blood characterization by smartphone colorimetry, the increases in M, Y and K values with time are analyzed by the linear least square fitting with R² listed in Table 2. In contrast to the implementation by Thanakiatkrai et al. (2013), the M value cannot be an essential parameter for blood characterization. The K value in Figure 7, corresponding to the darkness in the CMYK color space, is efficient in describing the blood changes. The K value correlates highly with bloodstain age, with R² values of 0.7877 on cloths and 0.9602 on papers, nearly matching those of the V value.

Figure 7. The variation of K value with the time since deposition of blood up to 11 days (t) from sample type, (a) A, (b) B, (c) AB and (d) O.

Table 2. R² from the linear least square fitting of RGB, HSV and MYK plots against time (t).

Plot		Bloodstain on cotton cloth			Bloodstain on tissue paper
X axis	Y axis	Maximum R^2	Average R^2	Minimum R^2	Maximum R^2	Average R^2	Minimum R^2
ln t	R	0.9822 (A)	0.6855	0.2146 (O)	0.9902 (A)	0.8414	0.7218 (AB)
t	R	0.9565 (A)	0.7957	0.4937 (O)	0.9840 (O)	0.9284	0.8217 (AB)
t	G	0.9830 (A)	0.7665	0.3547 (O)	0.9704 (A)	0.9058	0.7299 (O)
t	B	0.9816 (A)	0.8798	0.8054 (AB)	0.9972 (AB)	0.8970	0.7339 (O)
t	H	0.9381 (A)	0.3697	0.0100 (B)	0.5349 (B)	0.2084	0.0144 (A)
t	S	0.9167 (AB)	0.6359	0.1962 (O)	0.9906 (A)	0.7173	0.0715 (O)
t	V	0.9756 (AB)	0.8254	0.4760 (O)	0.9824 (O)	0.9602	0.9330 (A)
t	M	0.8291 (AB)	0.5371	0.1467 (O)	0.8293 (A)	0.6598	0.1429 (O)
t	Y	0.9158 (AB)	0.7703	0.5562 (B)	0.9632 (B)	0.7322	0.4578 (O)
t	K	0.9341 (A)	0.7877	0.4937 (O)	0.9840 (O)	0.9340	0.8217 (AB)

Consistent with the term ‘red blood cell’, the R value is a primary component of blood color due to oxygenated hemoglobin. After leaving the body, the blood exposes to oxygen, but the oxyhemoglobin transformation to its derivatives highlights the importance of monitoring a linear reduction in B with a large slope. The V value was suggested as the best parameter describing the color changes in the case of the constant components. In the case of temporal change of blood color, the V was exclusively used in Choi et al. (2019). This work confirms that the darker blood due to red blood cell degradation is suitably described by the reduction in V. The corresponding K parameter was not often used in the literature (Choi et al., 2019; Marrone et al., 2021; Shin et al., 2017), but exhibits a good sensitivity to the time since deposition in this study.

The results show that blood color monitoring is quantifiable using digital image colorimetry. Smartphone colorimetry is inherently sensitive to lighting, camera and colorimetric application (Shin et al., 2017; Thanakiatkrai et al., 2013). For instance, the V values of bloodstains are significantly higher when a high-intensity light source or a smartphone’s flashlight is applied. The methods of eliminating the environmental and instrumental effects have been proposed, as exemplified by Nixon, Outlaw, and Leung (2020). Because the measurement trend remains consistent, smartphone colorimetry is still promising for on-site analysis. However, a critical issue arises that colorimetric parameters in this study are significantly varied even with the white absorbent substrates. This finding suggests that the blood clumping observed on leather and glass substrates by Thanakiatkrai et al. (2013), also occurs on cloths and papers. It follows that an uneven layer leads to different light reflections. To increase the reliability of this system, the variations of R, B, V and K parameters with the thickness of bloodstain should be further explored.

Finally, the effect of different smartphones is discussed. Each smartphone camera gives rise to different color values depending on its image sensors and other settings. Images must be collected by the same smartphone for the entire process to justify the comparison of any parameter (Fan et al., 2021), without using an automatic white balance to avoid measurement deviations (Zamora-Garcia, Correa-Tome, Hernandez-Belmonte, Ayala-Ramirez, & Ramirez-Paredes, 2021). In this study, the temporal changes in colors were monitored by a single smartphone, and the variations in different color parameters were compared. If the absolute values are required, the smartphone-based measurement should be calibrated with a standard color chart or colorimeter under a control lighting condition. Also, a subsequent correction of color data has been suggested, as demonstrated in You, Liu, Zhang, Xv, and He (2020).

4. Conclusion

The bloodstains deposited on cloths and tissue papers for 1–11 days were imaged by a smartphone camera, and the RGB, HSV, as well as CMYK colorimetric parameters, were assessed by the ‘Colorimeter X’ application on the smartphone. The linear reductions in R, G, B and V values correspond to the color changes from bright red to dark brown. The average value of R² from the linear least square fitting was the highest in the case of the V value, making it the most appropriate parameter for blood monitoring. When the R and V were severely affected by measurement deviations, the B value emerged as a parameter to complement the R and V values in blood characterizations. The Y, M and K values increased during the blood ageing process, and the R² from linear correlations between K value and time were slightly lower than those in the case of V, revealing K as another effective parameter for blood characterization. These preliminary results show that this rapid, cost-effective and truly portable methodology can be considered for on-site analysis. Unlike the blood pattern analysis, digital image colorimetry can monitor the changes in blood deposition on absorbent substrates.

Acknowledgments

The authors are thankful for the advice and support from Sampart Cheedket, Maytinee Sriwilai, Wichuta Jakkan, Jantanipa Nuanjan, Natnichaphan Nualvijit, and Associate Professor Manas Kotepui.

Disclosure Statement

On behalf of all authors, the corresponding author states that there is no conflict of interest.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.