A new parameter enhancing breast cancer detection in computer-aided diagnosis of X-ray mammograms

DOI Web Site PubMed 1 Citations 19 References
  • Tanki Nobuyoshi
    Department of Medical Physics and Engineering, Division of Medical Technology and Science, Course of Health Science, Graduate School of Medicine, Osaka University Department of Clinical Radiology, Faculty of Health Sciences, Hiroshima International University
  • Murase Kenya
    Department of Medical Physics and Engineering, Division of Medical Technology and Science, Course of Health Science, Graduate School of Medicine, Osaka University
  • Nagao Michinobu
    Department of Radiology, Ehime Prefectural Imabari Hospital

Search this article

Abstract

The purpose of this study was to introduce a new parameter which enhances breast cancer detection using X-ray mammography. We used the database of X-ray mammograms generated by the Japan Society of Radiological Technology. The new parameter called 'quasi-fractal dimension (Q-FD)' was calculated from the relationship between the cutoff values for the maximum image intensity in the lesion set at 21levels from 20% to 100% at equal intervals and the number of pixels with an intensity exceeding the cutoff value. In addition to Q-FD, the image features such as curvature (C) and eccentricity (E) were extracted. The conventional fractal dimension (C-FD)was also calculated using the box-counting method. We used artificial neural networks (ANNs) as a classification method. When using C, E, C-FD and age as inputs in ANNs and taking the number of neurons in the hidden layer as 50, we found the area under the receiver operating characteristic curve (As) was 0.87 ± 0.07 in the task differentiating between benign and malignant masses. When Q-FD was added to inputs in addition to the above parameters, the As value was significantly improved to become 0.93 ± 0.09. These results suggested that Q-FD is effective for discriminating between benign and malignant masses.

Journal

Citations (1)*help

See more

References(19)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top