Local Time : 11:00 CET

C-0673 - Discrimination between metastatic and osteoporotic vertebral fractures with MRI using an evidence-based Bayesian network

M. Benndorf, T. Diallo, M. Langer, E. Kotter; Freiburg/DE Type: Scientific Exhibit
Area of Interests: Musculoskeletal spine, Bones
Imaging Technique: MR
Procedures: CAD, Computer Applications-Detection, diagnosis, Decision analysis
Special Focuses: Metastases, Osteoporosis
Add poster to my schedule In your schedule (remove)

Aims and objectives: Discrimination between osteoporotic vertebral fractures (OVF) and vertebral fractures due to metastasis (MVF) with MRI is a challenging radiological task. An accurate diagnosis is vital since suspicion for an MFV may require further workup. Furthermore, if the suspicion is false positive, it causes [...]

Methods and materials: Imaging features analyzed We employ the published diagnostic accuracies of 13 MR imaging features to differentiate benign OVF and malignant MVF [1], sensitivities and specificities for the features are given in table 1.   The decision help   To calculate a probability for MVF given a set[...]

Results: Using our classifier, 4/4 MVF and 14/17 OVF are classified correctly. Figures 4 and 5 demonstrate the mode of operation of our Bayes classifier – the decision help reports probabilities that favor the correct diagnosis.

Conclusion: We provide an interactive framework to distinguish metastatic from osteoporotic vertebral fractures with MRI using a naïve Bayesian approach. The strength of our work is the automatic integration of various diagnostic information (the accuracies of the single imaging features) to form a single proba[...]

Personal information: All authors: University Hospital Freiburg, Germany. Hugstetter Straße 55 79106 Freiburg, Germany Contact: matthias.benndorf[at]uniklinik-freiburg.de

References:   1.         Jung H-S, Jee W-H, McCauley TR, et al (2003) Discrimination of metastatic from acute osteoporotic compression spinal fractures with MR imaging. RadioGraphics 23:179–187. doi: 10.1148/rg.231025043   2.        [...]

This website uses cookies. Learn more