Accurate Fully Automatic Femur Segmentation in Pelvic Radiographs Using Regression Voting

Abstract

Abstract. Extraction of bone contours from radiographs plays an important role in disease diagnosis, pre-operative planning, and treatment analysis. We present a fully automatic method to accurately segment the proximal femur in anteroposterior pelvic radiographs. A number of candidate positions are produced by a global search with a detector. Each is then refined using a statistical shape model together with local detectors for each model point. Both global and local models use Random Forest regression to vote for the optimal positions, leading to robust and accurate results. The performance of the system is evaluated using a set of 519 images. We show that the fully automated system is able to achieve a mean point-to-curve error of less than 1mm for 98% of all 519 images. To the best of our knowledge, this is the most accurate automatic method for segmenting the proximal femur in radiographs yet reported.

3D statistical shape models incorporating 3D random forest regression voting for robust CT liver segmentation

“…In order to reduce training time, node-wise subsampling was performed if the node samples exceeded a count of 500. 4 Only one single high-resolution version was trained for the subsequent segmentation tasks. Training time for 2562 forests took 3 days on 5-7 intel hexacore computers.…”

“…A particular approach, the Random Forest Regression Voting, has been employed for the 2D SSM segmentation of the human femur on 2D radiographs, where excellent results were achieved. 4 Recently, we have presented for the first time a method for 3D SSM landmark appearance modeling based on 3D Random Forest Regression Voting, enabling highly robust omni-directional landmark detection. 5 In this work, we present first quantitative segmentation results for the human liver in CT volumes, based on 3D SSMs incorporating 3D Random Forest Regression Voting.…”

A complete-pelvis segmentation framework for image-free total hip arthroplasty (THA): methodology and clinical study

“…To our knowledge, the fast, robust and accurate segmentation of the complete pelvis in AP radiographs has not yet been fully explored, despite numerous methods developed and successful results achieved by previous works for the proximal femur delineation, due to the more complicated anatomical structures in the pelvic region than those in the femoral area.…”

“…Attempts to develop an automatic solution for the proximal femur segmentation have been presented previously . Despite numerous methods developed and successful results achieved by previous work for delineating the proximal femur, a fast and robust segmentation of the complete pelvis in AP radiographs with sufficient accuracy has not yet been fully explored due to more complicated anatomical structures in the pelvic region than those in the femoral area.…”

Development of a fully automatic shape model matching (FASMM) system to derive statistical shape models from radiographs: application to the accurate capture and global representation of proximal femur shape

“…As the current semi-automatic ASM methodology is timeconsuming and vulnerable to intra-and inter-operator error, its application to large-scale morphometric analyses is limited. We have therefore developed a fully automatic shape model matching (FASMM) system that uses an SSM to capture and represent the shape of skeletal element/s within standard radiographs 12 . In this paper, we demonstrate the use of the FASMM system for segmentation of the proximal femur from anteroposterior (AP) radiographs via its rapid and accurate placement of 65 reference points along the contour of the proximal femur without any manual intervention.…”