Sex estimation of the sternum by automatic image processing of multi-slice computed tomography images in a Croatian population sample: a retrospective study

Abstract

Aim: To determine the sexual dimorphism of the sternum with standard measurements in a contemporary Croatian population sample using multi-slice computed tomography (MSCT) and to compare the data obtained by an automatic with those obtained by a manual approach.

Methods: Five sternal measurements were obtained from MSCT images of 73 men and 55 women and three sternal indices were calculated. Custom image analysis software was developed for automatic segmentation and calculation of sternal measurements. Measurements of sexual dimorphism were automatically calculated and compared with manual measurements.

Results: All of the sternal measurements exhibited significant differences between men and women. The discrepancies between manual and automatic measurements ranged from 2.8% to 3.6% of the mean average values obtained with the automatic approach. The most accurate single-variable discriminant function was sternal body length (82.8%), the most accurate index was sternal area (89.1%), and the discriminant function using three variables was manubrium width, sternal body length, and sternal body width (90.6%).

Conclusion: Sternal measurements are a reliable sex indicator and can be used in forensic casework. Computer-aided measurement methods can accelerate sex estimation and improve its precision and accuracy.

Figure 1

Identification of contours and determination of manubrium (left panel) and sternal body (right panel) osteometric measurements by the automatic approach. Segmentation contours (yellow lines) and contour’s body center-of-mass position (central pink dots).

Figure 2

Distribution of pairwise differences between manual and automatic measurements for sternal body length (B) and sternal body width at level of the incisurae costales 1 (CSW1) and 2 (CSW2) (A) and manubrium length (M), manubrium width (MW) (B). Each panel has the same number of bins (n = 11), producing different bin widths across osteometric measurements.

Figure 3

The multi-slice computed tomography (MSCT) images of two sternal bodies showing exemplary situations for landmark recognition. MSCT scan on the left allowed easy identification of landmarks for sternal body length (B), while the right one has fewer clear contours.

Figure 4

Sexual dimorphism of sternal body length (B) and sternal body width at level of the incisurae costales 1 (CSW1) and 2 (CSW2) (A) and manubrium length (M), manubrium width (MW) (B). Male – blue, female – red.