Refine
Document Type
- In Proceedings (5) (remove)
Language
- English (5)
Has Fulltext
- no (5)
Is part of the Bibliography
- no (5)
Keywords
- statistical shape and intensity models (2)
- 3d-reconstruction from 2d X-rays (1)
- 3d-reconstruction from 2d Xrays (1)
- Articulated Models, Statistical Shape And Intensity Models, 2D/3D Anatomy Reconstruction, Pelvic Parameters Measurement, Total Hip Arthroplasty (1)
- fracture fixation of the distal femur (1)
- osteosynthesis follow-up (1)
- pose estimation (1)
Institute
Towards Robust Measurement Of Pelvic Parameters From AP Radiographs Using Articulated 3D Models
(2015)
Patient-specific parameters such as the orientation of the acetabulum or pelvic tilt are useful for custom planning for total hip arthroplasty (THA) and for evaluating the outcome of surgical interventions. The gold standard in obtaining pelvic parameters is from three-dimensional (3D) computed tomography (CT) imaging. However, this adds time and cost, exposes the patient to a substantial radiation dose, and does not allow for imaging under load (e.g. while the patient is standing). If pelvic parameters could be reliably derived from the standard anteroposterior (AP) radiograph, preoperative planning would be more widespread, and research analyses could be applied to retrospective data, after a postoperative issue is discovered.
The goal of this work is to enable robust measurement of two surgical parameters of interest: the tilt of the anterior pelvic plane (APP) and the orientation of the natural acetabulum. We present a computer-aided reconstruction method to determine the APP and natural acetabular orientation from a single, preoperative X-ray. It can easily be extended to obtain other important preoperative and postoperative parameters solely based on a single AP radiograph.
We present a method based on a generative model for detection of disturbances such as prosthesis, screws, zippers, and metals in 2D radiographs. The generative model is trained in an unsupervised fashion using clinical radiographs as well as simulated data, none of which contain disturbances. Our approach employs a latent space consistency loss which has the benefit of identifying similarities, and is enforced to reconstruct X-rays without disturbances. In order to detect images with disturbances, an anomaly score is computed also employing the Frechet distance between the input X-ray and the reconstructed one using our generative model. Validation was performed using clinical pelvis radiographs. We achieved an AUC of 0.77 and 0.83 with clinical and synthetic data, respectively. The results demonstrated a good accuracy of our method for detecting outliers as well as the advantage of utilizing synthetic data.
We present a method for the automated segmentation of knee bones and cartilage from magnetic resonance imaging, that combines a priori knowledge of anatomical shape with Convolutional Neural Networks (CNNs). The proposed approach incorporates 3D Statistical Shape Models (SSMs) as well as 2D and 3D CNNs to achieve a robust and accurate segmentation of even highly pathological knee structures. The method is evaluated on data of the MICCAI grand challenge "Segmentation of Knee Images 2010". For the first time an accuracy equivalent to the inter-observer variability of human readers has been achieved in this challenge. Moreover, the quality of the proposed method is thoroughly assessed using various measures for 507 manual segmentations of bone and cartilage, and 88 additional manual segmentations of cartilage. Our method yields sub-voxel accuracy. In conclusion, combining of anatomical knowledge using SSMs with localized classification via CNNs results in a state-of-the-art segmentation method.
We present a novel method to derive the surface distance of an osteosynthesis plate w.r.t. the patientspecific surface of the distal femur based on 2D Xray images. Our goal is to study from clinical data, how the platetobone distance affects bone healing. The patientspecific 3D shape of the femur is, however, seldom recorded for cases of femoral osteosynthesis since this typically requires Computed Tomography (CT), which comes at high cost and radiation dose. Our method instead utilizes two postoperative Xray images to derive the femoral shape and thus can be applied on radiographs that are taken in clinical routine for followup. First, the implant geometry is used as a calibration object to relate the implant and the individual Xray images spatially in a virtual Xray setup. In a second step, the patientspecific femoral shape and pose are reconstructed in the virtual setup by fitting a deformable statistical shape and intensity model (SSIM) to the images. The relative positioning between femur and implant is then assessed in terms of displacement between the reconstructed 3D shape of the femur and the plate. A preliminary evaluation based on 4 cadaver datasets shows that the method derives the platetobone distance with a mean absolute error of less than 1mm and a maximum error of 4.7 mm compared to ground truth from CT. We believe that the approach presented in this paper constitutes a meaningful tool to elucidate the effect of implant positioning on fracture healing.
We present a novel method to derive the surface distance of an osteosynthesis plate w.r.t. the patient-specific surface of the distal femur based on postoperative 2D radiographs. In a first step, the implant geometry is used as a calibration object to relate the implant and the individual X-ray images spatially in a virtual X-ray setup. Second, the patient- specific femoral shape and pose are reconstructed by fitting a deformable statistical shape and intensity model (SSIM) to the X-rays. The relative positioning between femur and implant is then assessed in terms of the displacement between the reconstructed 3D shape of the femur and the plate. We believe that the approach presented in this paper constitutes a meaningful tool to elucidate the effect of implant positioning on fracture healing and, ultimately, to derive load recommendations after surgery.