How random is the random forest ? Random forest algorithm on the service of structural imaging biomarkers for Alzheimer's disease: from Alzheimer's disease neuroimaging initiative(ADNI) database

Neuroinformatics is a fascinating research field that applies computational models and analytical tools to high dimensional experimental neuroscience data for a better understanding of how the brain functions or dysfunctions in brain diseases. Neuroinformaticians work in the intersection of neuroscience and informatics supporting the integration of various sub-disciplines（behavioural neuroscience, genetics, cognitive psychology, etc.） working on brain research. Neuroinformaticians are the pathway of information exchange between informaticians and clinicians for a better understanding of the outcome of computational models and the clinical interpretation of the analysis. Machine learning is one of the most significant computational developments in the last decade giving tools to neuroinformaticians and finally to radiologists and clinicians for an automatic and early diagnosis-prognosis of a brain disease. Random forest（RF） algorithm has been successfully applied to high-dimensional neuroimaging data for feature reduction and also has been applied to classify the clinical label of a subject using single or multi-modal neuroimaging datasets. Our aim was to review the studies where RF was applied to correctly predict the Alzheimer＇s disease（AD）, the conversion from mild cognitive impairment（MCI） and its robustness to overfitting, outliers and handling of non-linear data. Finally, we described our RF-based model that gave us the 1 ^（st） position in an international challenge for automated prediction of MCI from MRI data.

Brain Findings Associated with Iodine Deficiency Identified by Magnetic Resonance Methods: A Systematic Review

Objectives: Iodine deficiency (ID) is a common cause of preventable brain damage and mental retardation worldwide, according to the World Health Organisation. It may adversely affect brain maturation processes that potentially result in structural and metabolic brain abnormalities, visible on Magnetic Resonance (MR) techniques. Currently, however, there has been no review of the appearance of these brain changes on MR methods. Methods: A systematic review was conducted using 3 online search databases (Medline, Embase and Web of Knowledge) using multiple combinations of the following search terms: iodine, iodine deficiency, magnetic resonance, MRI, MRS, brain, imaging and iodine deficiency disorders (i.e. hypothyroxinaemia, congenital hypothyroidism, hypothyroidism and cretinism). Results: Up to May 2013, 1673 related papers were found. Of these, 29 studies confirmed their findings directly using MR Imaging and/or MR Spectroscopy. Of them, 28 were in humans and involved 157 subjects, 46 of whom had primary hypothyroidism, 97 had congenital hypothyroidism, 3 had endemic cretinism and 11 had subclinical hypothyroidism. The studies were small, with a mean relevant sample size of 6, median 2, range 1 - 35, while 14 studies were individual case reports. T1-weighted was the most commonly used MRI sequence (20/29 studies) and 1.5 Tesla was the most commonly used magnet strength (6/10 studies that provided this information). Pituitary abnormalities (18/29 studies) and cerebellar atrophy (3/29 studies) were the most prevalent brain abnormalities found. Only fMRI studies (3/29) reported cognition-related abnormalities but the brain changes found were limited to a visual description in all studies. Conclusions: More studies that use MR methods to identify changes on brain volume or other global structural abnormalities and explain the mechanism of ID causing thyroid dysfunction and hence cognitive damage are required. Given the role of MR techniques in cognitive studies, this review provides a starting point for researching the macroscopic structural brain changes caused by ID.

Color Fusion of Magnetic Resonance Images Improves Intracranial Volume Measurement in Studies of Aging

Background: Comparison of intracranial volume (ICV) measurements in different subpopulations offers insight into age-related atrophic change and pathological loss of neuronal tissue. For such comparisons to be meaningful the accuracy of ICV measurement is paramount. Color magnetic resonance images (MRI) have been utilised in several research applications and are reported to show promise in the clinical arena. Methods: We selected a sample of 150 older community-dwelling individuals (age 71 to 72 years) representing a wide range of ICV, white matter lesions and atrophy. We compared the extraction of ICV by thresholding on T2*-weighted MR images followed by manual editing (reference standard) done by an analyst trained in brain anatomy, with thresholding plus computational morphological operations followed by manual editing on a framework of a color fusion technique (MCMxxxVI) and two automatic brain segmentation methods widely used, these last three done by two image analysts. Results: The range of ICV was 1074 to 1921 cm3 for the reference standard. The mean difference between the reference standard and the ICV measured using the technique that involved the color fusion was 2.7%, while it was 5.4% compared with any fully automatic technique. However, the 95% confidence interval of the difference between the reference standard and each method was similar: it was 7% for the segmentation aided by the color fusion and was 7% and 8.3% for the two fully automatic methods tested. Conclusion: For studies of aging, the use of color fusion MRI in ICV segmentation in a semi-automatic framework delivered best results compared with a reference standard manual method. Fully automated methods, while fast, all require manual editing to avoid significant errors and, in this post-processing step color fusion MRI is recommended.