Differentiation of EMCI in sMR images using segmented brainstem multifractal texture measures

• Published in: Electronics letters Vol. 55; no. 23; pp. 1213 - 1214
• Main Authors: Palanisamy, R, Swaminathan, R, Sundar, S
• Format: Journal Article
• Language: English
• Published: The Institution of Engineering and Technology 14.11.2019
• Subjects: Alzheimer's condition; Bayes methods; biomarker imaging; biomedical MRI; Biomedical technology; brain; brainstem multifractal spectrum‐based texture measures; cognition; component labelling method; decision tree; decision trees; diseases; early mild cognitive impairment; extracted multifractal spectrum‐based texture measures; feature extraction; fractals; fuzzy C means‐based connected component labelling method; image classification; image segmentation; image texture; medical image processing; naive Bayes; neurophysiology; public domain database; random forest; segmented brainstem multifractal texture measures; statistical analysis; structural magnetic resonance images; biomedical MRI; image classification; segmented brainstem multifractal texture measures; extracted multifractal spectrum-based texture measures; random forest; component labelling method; brainstem multifractal spectrum-based texture measures; fractals; Alzheimer's condition; decision tree; image segmentation; feature extraction; medical image processing; cognition; early mild cognitive impairment; diseases; biomarker imaging; structural magnetic resonance images; brain; public domain database; image texture; fuzzy C means-based connected component labelling method; decision trees; naive Bayes; Bayes methods; neurophysiology; statistical analysis
• ISSN: 0013-5194; 1350-911X; 1350-911X
• DOI: 10.1049/el.2019.2821

In this Letter, an attempt has been made to segment the brainstem structure from structural magnetic resonance (sMR) images and differentiate early mild cognitive impairment (EMCI) from normal and Alzheimer's condition using multifractal texture measures. The images considered from public domain database are preprocessed and brainstem is segmented using fuzzy ‘C’ means-based connected component labelling method. Multifractal spectrum (MS) is evaluated for the segmented brainstem structure using the multifractal detrended moving average technique. Seven MS-based texture measures are extracted and statistically analysed using one-way analysis of variance. Classification algorithms namely, naïve Bayes, decision tree and random forest (RF) are employed to distinguish the EMCI condition. Ten-fold cross-validation approach is executed and its performance is evaluated using accuracy, precision, recall and area under the curve. The results indicate that the segmented brainstem structures possess multifractal characteristics. Of the extracted MS-based texture measures, αmax, α0, αR, B and Δα are found to be statistically significant. RF gives the highest accuracy of 92.3% in distinguishing EMCI from other subject groups. Hence, the proposed approach with brainstem MS-based texture measures in combination with RF can be used as an imaging biomarker for the diagnosis of EMCI.