Article Text

PDF
FRI0545 A Novel Feature Selection Algorithm Based on Bone Micro Architecture Analysis To Identify Osteoarthritis
  1. R. Ljuhar1,
  2. B. Norman1,
  3. D. Ljuhar1,
  4. T. Haftner1,
  5. J. Hladuvka2,
  6. M. Bui Thi Mai2,
  7. H. Canhão3,
  8. J. Branco3,
  9. A.M. Rodrigues3,
  10. N. Gouveia3,
  11. S. Nehrer4,
  12. A. Fahrleitner-Pammer5,
  13. H.-P. Dimai5
  1. 1Braincon Technologies
  2. 2VRVis Research Competence Center, Vienna, Austria
  3. 3Faculdade de Medicina da Universidade de Lisboa, Lisbon, Portugal
  4. 4Center for Regenerative Medicine & Orthopedics, Danube University, Krems
  5. 5Department of Internal Medicine, Division of Endocrinology and Metabolism, Medical University of Graz, Graz, Austria

Abstract

Background Texture information of the subchondral bone area of 2D radiographs represents a promising possibility for evaluating the state of osteoarthritis (OA) in addition to traditional clinical means such as visual and semi-quantitative assessments.

Objectives Algorithms based on fractal analysis have shown to be capable of identifying differences in trabecular bone structure. However such features are likely to vary within the subchondral bone area and therefore the appropriate selection of the region of interest (ROI) plays a crucial role for the result of the analysis. Thus, a feature selection algorithm is being applied in order to determine ROIs that enable an optimum discrimination between patients with and without OA.

Methods The study included 152 standardized knee radiographs from 66 female patients with OA, and 86 controls. Subchondral bone micro architecture was assessed by using both fractal analysis and a Shannon Entropy (SE) algorithm at predefined regions of the proximal tibia and the distal femur. For fractal analysis the distinct parameter Bone Structure Value (BSV) was defined. The selected area of the proximal tibia involved a matrix of 3x8 ROIs, whereas a 2x2 matrix was defined for each condyle of the distal femur. SE and the BSV were calculated for each of the 32 ROIs, respectively. Based on these 64 variables, a feature selection algorithm was applied to determine the variables that showed the best discrimination power between Case and Control subjects.

Results By combining the BSV and SE, the odds ratio increased significantly from 3.08 (95% CI: 1.78–5.30) to 14.82 (95% CI: 6.69–32.83) when using 15 features, and to 39.75 (95% CI: 15.41–102.51) based on 10 features. By using the selected 10 features the accuracy was found to be 0.86. This showed to be a significant improvement compared to the accuracy achieved when calculating a single mean value for the 3x8 ROIs of the proximal tibia alone (0.62 vs. 0.86).

Conclusions The application of a feature selection algorithm in accordance with the combination of the two texture analysis methods shows a significant improvement with respect to the discrimination power between subjects with and without OA. The high odds ratios confirm that reliable results can be achieved by combining the BSV and the SE. This novel algorithm for the assessment of bine micro architecture may not only be useful in osteoarthritis subjects but also for the early prediction and assessment of other degenerative bone diseases like osteoporosis and rheumatoid arthritis.

Disclosure of Interest R. Ljuhar Employee of: Braincon Technologies, B. Norman Employee of: Braincon Technologies, D. Ljuhar Employee of: Braincon Technologies, T. Haftner Employee of: Braincon Technologies, J. Hladuvka: None declared, M. Bui Thi Mai: None declared, H. Canhão: None declared, J. Branco: None declared, A. M. Rodrigues: None declared, N. Gouveia: None declared, S. Nehrer: None declared, A. Fahrleitner-Pammer: None declared, H.-P. Dimai: None declared

Statistics from Altmetric.com

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.