Evaluating the age of rib fractures through ^99mTc-MDP SPECT strategy based on a machine learning algorithm: a preliminary study

Rib fractures are among the most common injuries in traffic accidents. Accurate evaluation of rib fracture age is important as it assists forensic experts in reconstructing injury processes and provides objective technical support for subsequent judicial judgments. Routine diagnostic methods sometimes cannot meet forensic requirements for fracture-age estimation. Technetium-99m-methyldiphosphonate (^99mTc-MDP), a radiotracer administered intravenously, binds dynamically to hydroxyapatite involved in fracture healing. ^99mTc-MDP uptake increases in areas with abundant blood flow and active osteoblast repair at the fracture site, offering potential for estimating fracture age. In this study, ^99mTc-MDP uptake was measured using the semi-quantitative target-to-non-target ratio (T/NT) through single-photon emission computed tomography. A total of 151 rib fractures were included from 55 individuals. Fracture age ranged from 3 to 491 days and was divided into three groups (within 30 days, 30-60 days, and over 60 days). T/NT values differed significantly among groups using Mann-Whitney U tests (P < 0.0001). The correlation coefficient (R²) between fracture age and T/NT was 0.36 (P < 0.0001). Six machine-learning models were then constructed, among which the support vector machine model had the best performance with area under the curve (AUC) of 0.91, F1-score of 0.78, accuracy of 0.80, and recall of 0.79 when identifying the three groups in the independent testing set. This study offers a novel perspective for estimating rib fracture age using functional imaging, which may provide essential information for case reconstruction and serve as a useful complementary tool in forensic bone trauma analysis.