A Bayesian Framework for Automated Cardiovascular Risk Scoring on Standard Lumbar Radiographs

Abstract

We present a fully automated framework for scoring a patient's risk of cardiovascular disease (CVD) and mortality from a standard lateral radiograph of the lumbar aorta. The framework segments abdominal aortic calcifications for computing a CVD risk score and performs a survival analysis to validate the score. Since the aorta is invisible on X-ray images, its position is reasoned from 1) the shape and location of the lumbar vertebrae and 2) the location, shape, and orientation of potential calcifications. The proposed framework follows the principle of Bayesian inference, which has several advantages in the complex task of segmenting aortic calcifications. Bayesian modeling allows us to compute CVD risk scores conditioned on the seen calcifications by formulating distributions, dependencies, and constraints on the unknown parameters. We evaluate the framework on two datasets consisting of 351 and 462 standard lumbar radiographs, respectively. Promising results indicate that the framework has potential applications in diagnosis, treatment planning, and the study of drug effects related to CVD.

Automatic Location of Vertebrae on DXA Images Using Random Forest Regression

“…Our initialisation failure of 3.8% on L4-T9 appears comparable to the 2% on L4-L1 for the baseline data in [13], when scaled by the number of vertebrae. The final failure rate of 8.8% appears higher, but [13] deals only with the lumbar, whereas our additional failures are due to multiple severe fractures in the thoracic spine. The thoracic vertebrae are harder to segment because the vertebrae are closer together (especially when affected by disc disease), resulting in more edge confusions between neighbouring vertebrae; and there is more overlaying structure from the ribs and scapulae.…”

“…The overall mean error of 0.74mm compares well to other methods (e.g. [12], mean error 1.4mm, or [13], mean errors 1.22 or 1.34mm). Our initialisation failure of 3.8% on L4-T9 appears comparable to the 2% on L4-L1 for the baseline data in [13], when scaled by the number of vertebrae.…”

“…[12], mean error 1.4mm, or [13], mean errors 1.22 or 1.34mm). Our initialisation failure of 3.8% on L4-T9 appears comparable to the 2% on L4-L1 for the baseline data in [13], when scaled by the number of vertebrae. The final failure rate of 8.8% appears higher, but [13] deals only with the lumbar, whereas our additional failures are due to multiple severe fractures in the thoracic spine.…”

Automated scoring of aortic calcification in vertebral fracture assessment images

“…A range of successful techniques have been applied including inpainting, 7 shape guided classification and active contours, 8 and appearance models with Bayesian inference. 9 DXA imaging has an even lower radiation dose than plain radiography and can be used to reliably measure AAC. 4,10 DXA images contain less signal from soft tissues compared to plain radiographs, but are a lower resolution and contain more noise.…”

Improving SMC sampler estimate by recycling all past simulated particles

“…There are heuristic reasons for why this may be true in general: the annealing of the target and spread of samples over the support means that SMC is less likely to be become trapped in posterior modes. Moreover, the application of such technique has shown promising results in order to tackle challenging problem in the signal processing literature [8][9][10].…”