Parametrization of textural patterns in 123I-ioflupane imaging for the automatic detection of Parkinsonism

• Published in: Medical physics (Lancaster) 2014-01, Vol.41 (1), p.012502-n/a
• Main Authors: Martinez-Murcia, F. J., Górriz, J. M., Ramírez, J., Moreno-Caballero, M., Gómez-Río, M.
• Format: Article
• Language: English
• Subjects: 123I‐ioflupane; Analysis of texture; Automation; Biological material, e.g. blood, urine; Haemocytometers; Brain; computer aided diagnosis; Degenerative diseases (Alzheimer's, ALS, etc); Digital computing or data processing equipment or methods, specially adapted for specific applications; diseases; Entropy; feature extraction; Haralick texture features; Image analysis; image classification; Image data processing or generation, in general; Image Processing, Computer-Assisted - methods; Image quality; image texture; iodine compounds; Measuring radioactive content of objects, e.g. contamination (whole‐body counters G01T011/63); medical image processing; Multivariate analysis; neurophysiology; Nortropanes; Parkinson's disease; Parkinsonian Disorders - diagnostic imaging; Pattern Recognition, Automated - methods; Scintigraphy; sensitivity; single photon emission computed tomography; Single photon emission computed tomography (SPECT); Spatial analysis; support vector machines; Three dimensional image processing; Tomography, Emission-Computed, Single-Photon - methods; Vector calculus
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1118/1.4845115

Purpose: A novel approach to a computer aided diagnosis system for the Parkinson's disease is proposed. This tool is intended as a supporting tool for physicians, based on fully automated methods that lead to the classification of123I-ioflupane SPECT images. Methods: 123I-ioflupane images from three different databases are used to train the system. The images are intensity and spatially normalized, then subimages are extracted and a 3D gray-level co-occurrence matrix is computed over these subimages, allowing the characterization of the texture using Haralick texture features. Finally, different discrimination estimation methods are used to select a feature vector that can be used to train and test the classifier. Results: Using the leave-one-out cross-validation technique over these three databases, the system achieves results up to a 97.4% of accuracy, and 99.1% of sensitivity, with positive likelihood ratios over 27. Conclusions: The system presents a robust feature extraction method that helps physicians in the diagnosis task by providing objective, operator-independent textural information about123I-ioflupane images, commonly used in the diagnosis of the Parkinson's disease. Textural features computation has been optimized by using a subimage selection algorithm, and the discrimination estimation methods used here makes the system feature-independent, allowing us to extend it to other databases and diseases.