Results

The majority of the patients included in this study sustained injuries from road traffic accidents (RTAs), representing 81.98% of the patients, whereas the other type that formed the highest frequency was hanging injuries, accounting for 15.14% of the total study population. Firearm injuries and electric shock were uncommon and, therefore, could not be taken into consideration for comparative statistical purposes (Table 1). RTA injuries constituted the majority of cases in the present study. Hanging injuries represented the second largest category, whereas firearm and electrocution injuries were comparatively infrequent. The unequal group distribution should be considered during the interpretation of comparative statistical findings, and it limits broader generalizability.

Injury type | Frequency (n) | Percentage (%)
Road traffic accident (RTA) | 314 | 81.98
Hanging | 58 | 15.14
Firearm injury | 7 | 1.83
Electrocution | 4 | 1.04
Total | 383 | 100

Descriptive statistics of morphometric and gray-level texture parameters are summarized in Table 2. Substantial variability was observed in morphometric and texture-based parameters, indicating heterogeneity in injury morphology and image characteristics. Area and perimeter demonstrated broad ranges, reflecting variability in injury extent and geometry. Variation in gray-level SD suggests differences in tissue texture and pixel intensity distribution among injury categories.

Parameter | Mean ± SD | Median (IQR) | Minimum-maximum
Length (cm) | 7.53 ± 5.11 | 6.40 (3.2-10.5) | 0.50-25.00
Width (cm) | 2.66 ± 2.21 | 2.10 (1.0-3.8) | 0.10-10.00
Area (cm²) | 18.45 ± 24.86 | 9.80 (2.5-24.2) | 0.03-168.70
Perimeter (cm) | 18.01 ± 16.12 | 12.40 (4.8-24.5) | 0.50-86.00
Minimum gray value | 41.52 ± 23.41 | 38 (22-57) | 0-120
Maximum gray value | 254.82 ± 2.91 | 255 (254-255) | 240-255
Median gray value | 96.24 ± 18.33 | 94 (82-108) | 42-155
Standard deviation | 32.48 ± 12.76 | 30.5 (24-39) | 8-78
Solidity | 0.79 ± 0.11 | 0.81 (0.73-0.88) | 0.42-0.98

Comparative analysis of morphometric and gray-level parameters among injury groups is presented in Table 3. Statistically significant differences were found in length, width, area, perimeter, SD, and solidity. The highest F-value was for solidity (F=11.94, p<0.001), showing big differences in boundary regularity between injury groups. RTA injuries were larger in dimension than hanging and other injuries. The gray-level SD varied significantly, too, pointing to differences in tissue heterogeneity and image texture characteristics among the groups.

Parameter | RTA (mean ± SD) | Hanging (mean ± SD) | Other injuries (mean ± SD) | F-value | df | p-value
Length (cm) | 7.82 ± 5.22 | 6.11 ± 4.31 | 5.90 ± 4.84 | 3.21 | 2,380 | 0.041*
Width (cm) | 2.79 ± 2.18 | 2.08 ± 1.91 | 1.95 ± 1.70 | 3.07 | 2,380 | 0.048*
Area (cm²) | 20.62 ± 25.88 | 10.54 ± 14.63 | 8.22 ± 12.81 | 5.96 | 2,380 | 0.003*
Perimeter (cm) | 19.48 ± 16.92 | 11.94 ± 10.88 | 10.52 ± 11.10 | 5.17 | 2,380 | 0.006*
Median gray value | 97.84 ± 18.11 | 90.40 ± 17.02 | 88.33 ± 15.20 | 2.49 | 2,380 | 0.084
Standard deviation | 34.12 ± 12.91 | 27.08 ± 10.44 | 24.91 ± 9.62 | 4.56 | 2,380 | 0.011*
Solidity | 0.82 ± 0.09 | 0.71 ± 0.11 | 0.68 ± 0.10 | 11.94 | 2,380 | <0.001*

Assessment of multicollinearity among predictor variables using VIFs is shown in Table 4. The multicollinearity analysis showed that the predictor variables had acceptable correlations. All VIF values stayed below 5.0, so there was no severe multicollinearity. While the area and perimeter had somewhat higher VIFs, they were still within acceptable limits. This suggests the regression model is stable and good for further analyses.

Variable | VIF | Tolerance
Length | 2.81 | 0.356
Width | 2.42 | 0.413
Area | 4.18 | 0.239
Perimeter | 4.72 | 0.212
Standard deviation | 1.64 | 0.61

Results of exploratory logistic regression analysis are presented in Table 5. Exploratory binary logistic regression found that area, gray-level SD, and solidity predict injury differentiation. Solidity, specifically, had the strongest link-Wald χ² equals 7.61, odds ratio of 8.24, 95% CI of 1.82 to 37.24, and p of 0.006. This means that injuries with smoother edges are more likely to fall into certain categories. Overall, the model worked great, with an Omnibus χ² of 21.84, 3 degrees of freedom, and p less than 0.001, showing the variables helped a lot in classifying injuries. With a Nagelkerke R² of 0.318, about 31.8% of the variation in injury types can be chalked up to these factors. Also, the Hosmer-Lemeshow test was not significant (χ² = 6.12, p = 0.634), showing a well-calibrated model, while 79.6% overall accuracy showed good prediction.

Variable/model statistic | β Coefficient | SE | Wald χ² | Odds ratio (OR) | 95% CI | p-value
Area (cm²) | 0.021 | 0.01 | 4.13 | 1.02 | 1.001-1.04 | 0.042*
Standard deviation | 0.062 | 0.026 | 5.61 | 1.06 | 1.01-1.11 | 0.018*
Solidity | 2.11 | 0.765 | 7.61 | 8.24 | 1.82-37.24 | 0.006*
Overall model statistics
Omnibus test (χ²) | - | - | 21.84 | - | - | <0.001*
Degrees of freedom (df) | - | - | 3 | - | - | -
Nagelkerke R² | - | - | 0.318 | - | - | -
Hosmer-Lemeshow χ² | - | - | 6.12 | - | - | 0.634
Classification accuracy (%) | - | - | 79.6 | - | - | -

Diagnostic performance of selected parameters obtained through ROC analysis is summarized in Table 6. The ROC analysis showed that the combined predictive model worked well, with an AUC of 0.84 (95% CI: 0.77-0.90). Of the individual predictors, solidity was the best (AUC=0.81), followed by gray-level SD and area. All predictors demonstrated discriminatory performance significantly better than chance, backed by the statistically significant Z statistics. So, these quantitative morphometric and texture parameters could be useful tools to support forensic injury classification.

Parameter | AUC (95% CI) | Z Statistic | Sensitivity (%) | Specificity (%) | p-value
Solidity | 0.81 (0.73-0.88) | 7.82 | 79.4 | 76.2 | <0.001*
Standard deviation | 0.74 (0.66-0.82) | 5.61 | 71.2 | 69.8 | <0.001*
Area | 0.70 (0.61-0.79) | 4.12 | 68.4 | 65.7 | 0.001*
Combined model | 0.84 (0.77-0.90) | 8.93 | 82.1 | 78.4 | <0.001*