Results

Study selection

The literature search identified 246 records across databases and supplementary sources. After title and abstract screening, 12 full-text articles were assessed for eligibility. Six studies were excluded due to inappropriate design (n = 2), absence of relevant outcomes (n = 2), or potential population overlap (n = 2). Finally, six studies met the inclusion criteria and were incorporated into the systematic review and quantitative analyses [31–36]. The study selection process is presented in Fig. 1.

Fig. 1: Flow diagram of study selection

Characteristics of the included studies

All six included studies applied micro-CT to detect or characterize GSR. Five studies evaluated biological tissues obtained from human lower limbs [31–35], whereas one study analyzed non-biological textile substrates [36].

Biological samples

All five studies involving biological substrates were conducted in Italy between 2011 and 2016 [31–35] and employed standardized human lower-limb sections obtained from surgical amputations. Each specimen measured approximately 6 cm in length and originated from male donors aged 20–50 years. Sample sizes ranged from 24 to 75 segments per study.

Experimental protocols were highly standardized across investigations. Gunshots were delivered using a semi-automatic Beretta Model 81 pistol (caliber 7.65 × 17 mm Browning, 0.32 ACP) with full metal jacket ammunition from the same manufacturing lot. The firearm was mounted on a fixed stand positioned approximately perpendicular to the target surface to ensure consistent trajectory. Depending on the study design, between 6 and 30 shots were fired per distance, with firing distances ranging from 5 to 40 cm.

All studies employed the same high-resolution micro-CT system (Skyscan 1172) under identical acquisition settings. Image reconstruction and analysis were performed using N-Recon, CT-An, and CT-Vox software. System calibration was conducted with a water phantom, and GSR particles were identified using a radiodensity threshold exceeding 1000 Hounsfield Units (HU). Detailed technical specifications are provided in Tables 1 and 2. Across studies, the primary quantitative outcome was the proportion of GSR detected as a function of shooting distance, summarized in Fig. 2.

Author | Giovanni Cecchetto et al. | Giovanni Cecchetto et al. | Paolo Fais et al. | Paolo Fais et al. | Chiara Giraudo et al. | Zuzanna Brożek-Mucha et al.
Year | 2011 | 2012 | 2013 | 2015 | 2016 | 2020
Country | Italy | Italy | Italy | Italy | Italy | Poland
Sample | 60 sections of human legs(~ 6 cm length each) | 60 sections of human legs(~ 6 cm length each) | 24(~ 6 cm length each) | 30(~ 6 cm length each) | 75(~ 6 cm length each) | 6 white cotton fabric fragments(~ 40 × 40 cm)
Sex | Male | 60 | 60 | 24 | 30 | 75 | NA
Female | 0 | 0 | 0 | 0 | 0 | NA
Age range | 20–50 | 20–50 | 20–50 | 20–50 | 20–50 | NA
Firearm | Semi-automatic PistolBeretta mod. 81 | Semi-automatic PistolBeretta mod. 81 | Semi-automatic PistolBeretta mod. 81 | Semi-automatic PistolBeretta mod. 81 | Semi-automatic PistolBeretta mod. 81 | Hunting rifle
Caliber | 7.65 × 17 mm Browning(or 0.32 ACP) | 7.65 × 17 mm Browning(or 0.32 ACP) | 7.65 × 17 mm Browning(or 0.32 ACP) | 7.65 × 17 mm Browning(or 0.32 ACP) | 7.65 × 17 mm Browning(or 0.32 ACP) | 12-gauge
Bullet characteristics | FMJ(0.32 ACP or 7.65 × 17 mm Browning SR), same lot | FMJ(0.32 ACP or 7.65 × 17 mm Browning SR), same lot | FMJ(0.32 ACP or 7.65 × 17 mm Browning SR), same lot | FMJ(0.32 ACP or 7.65 × 17 mm Browning SR), same lot | FMJ(0.32 ACP or 7.65 × 17 mm Browning SR), same lot | 12/70 W8MP cartridges; lead projectile, Brenneke type, 16.5 mm, 29.5 g; plastic wad 8.3 mm
Firearm setup | Positioned on fixed stand, perpendicular to sample | Positioned on fixed stand, perpendicular to sample | Positioned on fixed stand, perpendicular to sample | Positioned on fixed stand, perpendicular to sample | Positioned on fixed stand, perpendicular to sample | NR
Gunshots per distance | 10 | 20 | 6 | 30 | 25 | 1
Firing distances (cm) | 5, 15, 23, 30, and 40 | 5, 15, and 30 | 5, 15, and 30 | 15 | 5, 15, and 30 | 0, 10, 20, 30, 50, and 70
Analysis / Outcome | Detection and characterization of GSR; 3D reconstruction of gunshot wounds; quantification of GSR percentage in relation to firing distance. | Detection and characterization of GSR by micro-CT; comparison between fresh and composed samples (15 days at open air with an average temperature of 24.5 °C and na average humidity of 64.9%). | Detection and quantification of GSR in charred samples (4 min at a temperature of 600 °C) by micro-CT; comparison with fresh controls and stab wounds. | Evaluation of the effects of different conditions (fresh, fabric-covered, water-immersed for 1 day at 28 °C, decomposed for 15 days at open air with an average temperature of 24.5 °C and humidity of 64.9%, and heat-exposed inside a wood-burning stove for 4 min at 600 °C) on GSR detection. | Analysis of the impact of different fabrics (cotton, denim, leather, and nylon), and firing distance on GSR detection using micro-CT. | Micro-CT: It was necessary tocut the fragments of cotton woven fabric to dimensions 5 × 5 cm with a bullet holein the centre.

Author | Giovanni Cecchetto et al. | Giovanni Cecchetto et al. | Paolo Fais et al. | Paolo Fais et al. | Chiara Giraudo et al. | Zuzanna Brożek-Mucha et al.
Year | 2011 | 2012 | 2013 | 2015 | 2016 | 2020
Micro-CT model | Skyscan 1172 h Micro-CT(Skyscan, Artselaar, Belgium) | Skyscan 1172 h Micro-CT(Skyscan, Artselaar, Belgium) | Skyscan 1172 h Micro-CT(Skyscan, Artselaar, Belgium) | Skyscan 1172 h Micro-CT(Skyscan, Artselaar, Belgium) | Skyscan 1172 h Micro-CT(Skyscan, Artselaar, Belgium) | Nanotom 180 N(Ge Sensing & Inspection Technologies phoenix-ray Gmbh, wunstorf, Germany)
Voltage (kV) | 100 | 100 | 100 | 100 | 100 | 100
Current (µA) | 100 | 100 | 100 | 100 | 100 | 150
Filter | Al, 1 mm of thickness | Al, 1 mm of thickness | Al, 1 mm of thickness | Al, 1 mm of thickness | Al, 1 mm of thickness | NR
Voxel size (µm) | 13 | 13 | 13 | 13 | 13 | 20
FOV (pixels) | 1280 × 1024 | 1280 × 1024 | 1280 × 1024 | 1280 × 1024 | 1280 × 1024 | 2300 × 2300(Hamamatsu detector)
Scan rotation | 360º, step 0.4º,frame averaging 2 | 360º, step 0.4º,frame averaging 2 | 360º, step 0.4º,frame averaging 2 | 360º, step 0.4º,frame averaging 2 | 360º, step 0.4º,frame averaging 2 | 360º in 2400 steps,frame averaging 3, image skip 1
Scan duration | NR | NR | 45 min | NR | 45 min | 100 min (500 ms exposure)
VOI | 1 cm x 1 cm x 3.8 mm,centered on the wound | 1 cm x 1 cm x 3.8 mm,centered on the wound | 1 cm x 1 cm x 3.8 mm,centered on the wound | 1 cm x 1 cm x 3.8 mm,centered on the wound | 1 cm x 1 cm x 3.8 mm,centered on the wound | 5 × 5 cm,centered on the wound
Software | Reconstruction | N-Recon(Skyscan, Artselaar, Belgium) | N-Recon(Skyscan, Artselaar, Belgium) | N-Recon(Skyscan, Artselaar, Belgium) | N-Recon(Skyscan, Artselaar, Belgium) | N-Recon(Skyscan, Artselaar, Belgium) | DatosX 2.1.0 (GE) using Feldkamp algorithm
Analysis | CT-An(Skyscan, Artselaar, Belgium) | CT-An(Skyscan, Artselaar, Belgium) | CT-An(Skyscan, Artselaar, Belgium) | CT-An(Skyscan, Artselaar, Belgium) | CT-An(Skyscan, Artselaar, Belgium) | VGStudio Max 2.1 (Volume Graphics Gmbh), and Fiji
3D Rendering | CT-Vox(Skyscan, Artselaar, Belgium) | CT-Vox(Skyscan, Artselaar, Belgium) | CT-Vox(Skyscan, Artselaar, Belgium) | CT-Vox(Skyscan, Artselaar, Belgium) | CT-Vox(Skyscan, Artselaar, Belgium) | ImageJ (BoneJ plug-in)
HU calibration | Water phantom | Water phantom | Water phantom | Water phantom | Water phantom | NR
GSR threshold (HU) | > 1000 | > 1000 | > 1000 | > 1000 | > 1000 | NR

Fig. 2: Percentage of GSR detected by micro-CT according to shooting distance in studies using biological samples

Non-biological samples

The single study investigating non-biological substrates was conducted in Poland and published in 2020 [36]. The experimental material consisted of six standardized cotton fabric fragments measuring approximately 40 × 40 cm. Gunfire exposure was performed using a 12-gauge hunting rifle with 12/70 W8MP cartridges loaded with lead Brenneke projectiles. One shot was fired at each distance, and shooting distances ranged from direct contact (0 cm) to 70 cm (0, 10, 20, 30, 50, and 70 cm), providing a broader spatial gradient than that used in the biological models.

Micro-CT imaging was conducted using a Nanotom 180 N system. Reconstruction employed DatosX 2.1.0 with the Feldkamp algorithm, and image processing and particle analysis were performed using VGStudio Max, Fiji, and ImageJ. Detailed acquisition and reconstruction parameters are provided in Tables 1 and 2. Unlike the biological studies, which reported detection percentages, this investigation quantified GSR deposition based on particle size metrics. At contact shots (0 cm), inorganic GSR particles and aggregates ranged from 53.8 to 465.9 μm, with a mean particle size of 184.8 μm (SD = 77.8 μm). At 10 cm, a marked reduction in particle size was observed (40.0–240.0 μm), with a mean of 83.4 μm (SD = 38.0 μm). At longer distances, mean particle sizes increased again, reaching 150.8 μm (SD = 64.9 μm) at 50 cm and 186.0 μm (SD = 82.7 μm) at 70 cm. These findings indicate a non-linear relationship between shooting distance and mean GSR particle size in cotton fabric substrates.

Risk of bias assessment

Most studies provided detailed descriptions of experimental procedures and exposure measurements. However, strategies to control potential confounding factors were rarely specified. Variability in shooting conditions was limited primarily to firing distance, with minimal exploration of alternative firearms or ammunition types. Additionally, eligibility criteria were insufficiently reported in the non-biological study.

Based on the predefined assessment tool, studies achieved high proportions of positive methodological ratings; nevertheless, important sources of uncertainty remained. Consequently, the overall risk of bias was judged as low-to-moderate, with limitations related to restricted external validity and experimental standardization. The complete assessment is presented in Table 3.

Question | Cecchetto2011 | Cecchetto2012 | Fais2013 | Fais2015 | Giraudo2016 | Brożek-Mucha2020 | Total (%)
Were the criteria for inclusion in the sample clearly defined? | ▲ | ▲ | ▲ | ▲ | ▲ | ▬ | ▲ | 83.3
▬ | 16.7
Were the study subjects and the setting described in detail? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Was the exposure measured in a valid and reliable way? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Were objective, standard criteria used for measurement of the condition? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Were confounding factors identified? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Were strategies to deal with confounding factors stated? | ▬ | ▬ | ▬ | ▬ | ▬ | ▬ | ▲ | 0.0
▬ | 100.0
Were the outcomes measured in a valid and reliable way? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Was appropriate statistical analysis used? | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | ▲ | 100.0
▬ | 0.0
Total (%) | ▲ 87.5 | ▲ 87.5 | ▲ 87.5 | ▲ 87.5 | ▲ 87.5 | ▲ 75.0 | ▲ | 85.41
▬ 12.5 | ▬ 12.5 | ▬ 12.5 | ▬ 12.5 | ▬ 12.5 | ▬ 25.0 | ▬ | 14.59

Qualitative and quantitative synthesis

Qualitative assessment across the included studies confirmed the capability of micro-CT for non-destructive visualization and spatial localization of GSR. Particles were consistently identified as high-contrast radiodense voxels (typically > 1000 HU), enabling three-dimensional mapping of residue clusters within the substrate volume. In biological specimens, micro-CT allowed localization of particles embedded within subcutaneous layers, whereas in textile substrates it allowed visualization of particles trapped between fibers.

All quantitative analyses were based on four studies evaluating biological substrates exposed to gunfire at multiple distances [31–33, 35]. These investigations reported mean percentages of GSR detection according to shooting distance under different experimental conditions, including fresh, decomposed, and charred tissues, as well as fabric-covered surfaces (cotton, jeans, leather, and nylon). Extracted summary measures (means, standard deviations, sample sizes, and corresponding distances) are provided in Supplementary Table 1.

Random-effects meta-analyses

The overall random-effects model yielded a pooled mean GSR detection rate of 0.060% (95% CI: 0.016–0.104; p = 0.007). Sensitivity analyses showed only minor changes in pooled estimates following sequential exclusion of the 23 cm and 40 cm distance conditions. Exclusion of the 23 cm condition resulted in a pooled estimate of 0.061% (95% CI: 0.015–0.107; p = 0.009), while exclusion of the 40 cm condition yielded an estimate of 0.063% (95% CI: 0.017–0.109; p = 0.007). Simultaneous exclusion of both distance conditions produced a pooled estimate of 0.064% (95% CI: 0.016–0.112; p = 0.009). Compiled sensitivity analysis results are presented in Supplementary Table S2. Distance-stratified analyses showed decreasing detection with increasing firing distance, from 0.133% at 5 cm to 0.019% at 30 cm. Effect estimates were statistically significant at 5 cm and 30 cm (p < 0.05) but not at 15 cm (p = 0.124). Between-study heterogeneity was extremely high across all models (I² > 99%), indicating substantial between-study variability. Summary estimates and heterogeneity statistics are reported in Table 4, with forest plots presented in Figs. 3 and 4.

Meta-analysis | k | Estimated mean [%, 95% CI] | p-value | τ | τ² | I² (%) | Cochran’s Q testQ (df) | p(Q)
Global | 23 | 0.060 [0.016–0.104] | 0.007 | 0.107 | 0.012 | 99.99 | Q (22) = 6562.21 | < 0.001
5 cm | 7 | 0.133 [0.006–0.259] | 0.039 | 0.169 | 0.029 | 99.99 | Q (6) = 1404.34 | < 0.001
15 cm | 7 | 0.043 [-0.012–0.098] | 0.124 | 0.074 | 0.005 | 99.89 | Q (6) = 978.16 | < 0.001
30 cm | 7 | 0.019 [0.004–0.035] | 0.016 | 0.021 | < 0.001 | 99.78 | Q (6) = 3628.93 | < 0.001

Fig. 3: Global random-effects meta-analysis showing the estimated mean percentage of GSR detected by micro-CT, with corresponding 95% confidence intervals

Fig. 4: Distance-stratified random-effects meta-analyses showing the estimated mean percentages of GSR detected by micro-CT at different firing distances: (A) 5 cm, (B) 15 cm, and (C) 30 cm, each with corresponding 95% confidence intervals. The 23 cm and 40 cm conditions were not included because they were represented by a single study

Quadratic mixed-effects meta-regression

Mixed-effects meta-regression identified a significant nonlinear association between shooting distance and GSR detection. The quadratic term for the centered distance variable was statistically significant (p < 0.001), while the linear term was not statistically significant (p = 0.286). The negative quadratic coefficient indicated a concave (downward-opening) relationship between distance and detection rate.

The intercept corresponded to a predicted baseline detection of 0.033% (p < 0.001). Residual heterogeneity remained extremely high (I² = 99.52%), and distance accounted for a negligible proportion of the between-study variance (pseudo-R² = 0%), consistent with the substantial variability expected in forensic sampling conditions. The fitted curve estimated a maximum predicted detection at 17.79 cm (bootstrap 95% CI: 17.47–18.10 cm). Full regression coefficients, standard errors, and confidence intervals are presented in Table 5, and model predictions are illustrated in Fig. 5.

Term | Estimate(b) | Std. Error | z-value | p-value | τ² (total) | I² (total, %) | Pseudo-R²(%) | Apex [95% CI](cm) | QE(residual heterogeneity) | QM (test of moderators)
Intercept | 0.0330 | 0.0074 | 4.475 | < 0.0001 | 0.0004 | 99.52 | 0 | 17.79[17.47–18.10] | 6098.76 (df = 20)p < 0.0001 | 674.808 (df = 2)p < 0.0001
Firing distance(centered) | –0.0000 | 0.0000 | -1.067 | 0.2585
Firing distance² | –0.0001 | 0.0000 | –25.487 | < 0.0001

Fig. 5: Quadratic mixed-effects meta-regression of gunshot distance versus percentage of GSR detected by micro-CT, with fitted curve and 95% confidence intervals

Gaussian fits with bootstrap confidence intervals

Weighted Gaussian models were fitted separately for each substrate condition to describe the relationship between shooting distance and GSR detection. Parameter estimates (β₀, β₁) and bootstrap 95% confidence intervals are presented in Supplementary Table S3, with fitted curves shown in Fig. 6, where descriptive R² values are provided to indicate the degree of agreement between observed data and the fitted profiles. Fresh biological tissues exhibited the highest baseline detection, with β₀ = 0.46% (95% CI: 0.20–0.65) and a decay parameter β₁ = 15.94 cm (95% CI: 14.14–18.80). Decomposed and charred tissues showed substantially lower baseline values (β₀ = 0.03%), accompanied by larger uncertainty and wider confidence intervals for β₁. For textile substrates (cotton, jeans, leather, and nylon), baseline detection remained low (β₀ range: 0.003–0.049%), with large β₁ estimates, in some cases exceeding 100 cm, indicating minimal observable decay within the evaluated distance range. These estimates demonstrate marked variability in detection profiles across substrate conditions.

Fig. 6: Weighted Gaussian fits of GSR detection for each sample stage showing the parameters β0 and β1 with 95 percent bootstrap confidence intervals. R² values are displayed for descriptive purposes only

Descriptive R² values indicated a close agreement between observed data and the fitted Gaussian profiles for fresh biological tissue (R² = 0.99) and cotton (R² = 0.94). In contrast, negative R² values were observed for decomposed, charred, jeans, leather, and nylon substrates, indicating that the Gaussian model did not reduce residual variability relative to a constant model for these conditions.