The Assessment Of The Segment And Measurement Of Femoral Shape From X-ray Film Using Active Shape Model

Objective:To assess the effect of the segment of the femoral shape from the X-ray film using ASM.To compare the accuracy and the reliability of the measurements using ASM with manual method and give the evidence of the potential of the application of ASM in clinic.Materials and Methods:50standard pelvic radiographs were collected from the radiology department.Then the format of original radiographs was converted from DICOM to JPG. The resolution of the target images ranges from766x768to791×772.1.20images were randomly selected to create a training set,Five different models were created using different numbers of landmark points.Each of the20training images was loaded in turn,and the trained model automatically searched for the outline of the femur.Then the point-to-point error and point-to-line error between the original, manually generated points and those automatically generated by ASM was calculated for each model to assess the effect of the segmentation.This process enabled the best model to be chosen.2. The remaining30images were analyzed by the best model.After the segmentation,five parameters were measured include the diameter of the femoral head(R),the narrowest distance of the femoral neck(D),the neck-shaft angle,Offset and the length of the femoral neck(CL) using ASM and manual method by three observers.The time spent making the measurements was recorded using a stop watch.3. The intra-clAss correlation coefficient(ICC)was calculated for each dataset to assess the reliability;a two-way mixed model was used with statistical significance set at95%.Single measures was reported for the intra-observer calculations.The average ICC values for different methods were compared using t-test.Paired t-tests were used to examine the correlation between the measurement values using the two methods.Values were defined as statistically significant for all tests if p<0.05.Result:1. The best model in our study includes331andmark points for the model containing femoral shaft and25for the model without femoral shaft.The accuracy of the model without femoral shaft is better than the model with shaft.2. The intra-obserber reliability ICC results were on average10.25%higher (range-0.4%-65.2%) for all the ASM datasets compared with the manual data and the difference wasn’t statistically significant(Paired t-test,p=0.25).The confidence intervals(CI)were smaller for all ASM measurements compared with the manual measurements.The inter-obserber reliability ICC results showed similar trends to the intra-observer results.3. The results from the paired t-test for each parameter demonstrated good correlation between the measurements values from the two methods.There is no statistically significant difference between the two measurement methods.On average,the ASM method took117s to take the measurements(standard deviation0.13s),the manual measurements took125s on average(standard deviation1.16s).The difference is statistically significant(paired t-test,p<0.5).Conclusions:ASM can not only segment femoral shape from the standard pelvic radiographs but also can do the measurements of the parameters of the femur automatically.The accuracy and reliability of the ASM measurement method is similar to the manual method,but it can improve the efficiency of the measurement process.The automatical method has potential application values for the future clinical work.