Investigating machine learning techniques for detection of depression using structural MRI volumetric features

Structural MRI offers anatomical details and high sensitivity to pathological changes. It can demonstrate certain patterns of brain changes present at a structural level. Research to date has shown that volumetric analysis of brain regions has importance in depression detection. However, such analysis has had very minimal use in depression detection studies at individual level. Optimally combining various brain volumetric features/attributes, and summarizing the data into a distinctive set of variables remain difficult. This study investigates machine learning algorithms that automatically identify relevant data attributes for depression detection. Different machine learning techniques are studied for depression classification based on attributes extracted from structural MRI (sMRI) data. The attributes include volume calculated from whole brain, white matter, grey matter and hippocampus. Attributes subset selection is performed aiming to remove redundant attributes using three filtering methods and one hybrid method, in combination with ranker search algorithms. The highest average classification accuracy, obtained by using a combination of both SVM-EM and IG-Random Tree algorithms, is 85.23%. The classification approach implemented in this study can achieve higher accuracy than most reported studies using sMRI data, specifically for detection of depression.

An open source approach for regional cortical bone mineral density analysis

Objective: Cortical porosity, particularly at the endocortical region, is recognised to play a central role in the pathogenesis of bone fragility. Therefore, the purpose of this study was to: 1) demonstrate how cortical volumetric BMD (vBMD) distribution can be analysed from (p)QCT images and 2) highlight the clinical significance of assessing regional density distribution of cortical bone. 

Methods: We used pQCT to compare mid-tibial cortical volumetric BMD distribution of 20 young (age 24(SD2) years, mass 77(11) kg, height 178(6) cm) and 25 elderly (72(4) years, 75(9) kg, 172(5) cm) men. Radial and polar cortical vBMD distributions were analysed using a custom built open source analysis tool which allowed the cortex to be divided into three concentric cortical divisions and in 36 cortical sectors originating from the centroid of the bone.

Results: Mean vBMD did not differ between the groups (1135(16) vs. 1130(28) mg/cm, P=0.696). In contrast, there was a significant age-group by radial division interaction for radial cortical vBMD (P<0.001).

Conclusions: The proposed analysis method for analysing cortical bone density distribution of pQCT images was effective for detecting regional differences in cortical density between young and elderly men, which would have been missed by just looking at mean vBMD values.

An algorithm for determination of rank and degree of contribution of sMRI volumetric features in depression detection

Brain volume changes at structural level appear to have utmost importance in depression biomarkers studies. However, these brain volumetric findings have very minimal utilization in depression detection studies at individual level. Thus, this paper presents an evaluation of volumetric features to identify the relevant/optimal features for the detection of depression. An algorithm is presented for determination of rank and degree of contribution (DoC) of structural magnetic resonance imaging (sMRI) volumetric features. The algorithm is based on the frequencies of each feature contribution toward the desired accuracy limit. Forty-four volumetric features from various brain regions were adopted for evaluation. From DoC analysis, the DoC of each volumetric feature for depression detection is calculated and the features that dominate the contribution are determined.