Assessment of performance improvement in content-based medical image retrieval schemes using fractal dimension

The aim of this study was to investigate whether using a fractal dimension as an objective index (quantitative measure) to assess and control the "visual" or "texture" similarity of reference-image regions selected by a content-based image retrieval (CBIR) scheme would (or would not) affect the performance of the scheme in classification between image regions depicting suspicious breast masses.

An image data set depicting 1500 verified mass regions and 1500 false-positive mass regions was used. Fourteen morphologic and intensity distribution features and a fractal dimension were computed. A CBIR scheme using a k-nearest neighbor classifier was applied, and two experiments were conducted. In the first experiment, the CBIR scheme was evaluated using all 15 features. In the second experiment, the fractal dimension was used as a prescreening feature to guide the CBIR scheme to search for the most similar reference images that had similar measures in the fractal dimension.

The CBIR scheme achieved classification performance with areas under the receiver-operating characteristic curve of 0.857 (95% confidence interval [CI], 0.844-0.870) using 14 features and 0.866 (95% CI, 0.853-0.879) after adding the fractal dimension (P = .005 for both results). After using the fractal dimension as a prescreening feature, the CBIR scheme achieved an area under the receiver-operating characteristic curve of 0.851 (95% CI, 0.837-0.864), without a significant difference from the previous result using the original 14 features (P = .120). The difference of fractal dimension values between the selected similar reference images was reduced by 56.7%, indicating improvement in image texture similarity. In addition, more than half of references were discarded early, without similarity comparisons, indicating improvement in searching efficiency.

This study demonstrated the feasibility of applying a fractal dimension as an objective (quantitative) and efficient search index to assess and maintain the texture similarity of reference mass regions selected by a CBIR scheme without reducing the scheme's performance in classifying suspicious breast masses.