Ardekani Lab

The main focus of research by Dr. Ardekani’s group is on magnetic resonance image analysis and processing with applications to neuropsychiatric and neurological disorders. Developments center on four broad areas: image registration; applications of automatic pattern recognition techniques to MRI; modeling and estimation of water diffusion process in brain; and analysis of functional MRI and diffusion tensor imaging data. There is considerable overlap between these areas – for example, image registration results can be used to design better automated pattern recognition algorithms, and vice versa. The group’s main areas of application have traditionally been in Alzheimer’s disease (AD) and schizophrenia.

Current Investigations

Machine learning for prediction of conversion from mild cognitive impairment to Alzheimer’s Disease dementia
Multi-atlas methods for segmentation of the corpus callosum mid-sagittal cross-sectional area with application to characterizing Alzheimer’s disease progression
Measurement of hippocampal volumetric integrity with applications to characterizing schizophrenia and Alzheimer’s disease progression
Automated brain landmarks identification and detection methods with application to brain image registration and segmentation
Investigation into the functional and structural neuronal substrates of alcohol use disorders
Application of cloud computing for prediction of health outcomes

Specific Applications

Image Registration

Image registration is a fundamental step in medical image analysis arising in multiple applications. The general problem is to find a mathematical transformation that maps to each other pairs of anatomically corresponding points in the fields of view (FOVs) of independently acquired scans. Scans may come from the same subject (intra-subject registration) or different subjects (inter-subject registration); the same imaging modality (e.g., MRI-MRI registration) or different modalities (e.g., MRI-PET registration); and transformations can be linear or non-linear with different degrees of complexity. The Automatic Registration Toolbox (ART) developed by Dr. Ardekani’s group includes software designed to automatically find the registration transformation in several specific problems. These methods have been heavily used by the medical imaging research community. In a comparison of 14 non-linear registration algorithms published in 2009, ART was found to be in the top two methods in all comparison criteria.

Corpus Callosum

The corpus callosum is the largest white matter tract in the brain, interconnecting the left and right hemispheres. In spite of its significance, its unusual shape has led to controversies regarding its morphological characteristics and their relevance. Dr. Ardekani’s group has introduced methodological improvements to characterizing corpus callosum morphology including the introduction of a new metric, circularity, into this field of study. Software ‘yuki’ for measurement of corpus callosum morphology has been developed by the group and made publicly available to the research community. Several papers have been published in high-impact journals by researchers utilizing this software. ‘yuki’ is currently believed to be the best corpus callosum segmentation software publicly available in terms of speed, accuracy, robustness, reliability and ease of use.

Automated Diagnosis

While considerable obstacles still remain towards a practical computer-based system for diagnosis of psychiatric disorders, the group has applied pattern recognition and machine learning methods for automated diagnosis of schizophrenia based on the patterns of water diffusion in the brain measure using diffusion tensor imaging. After training the algorithm on 50 cases with known diagnoses (25 schizophrenia and 25 normal controls), the algorithm was presented with 50 new cases without specifying the diagnoses. The program was able to classify 49 cases correctly. Dr. Ardekani’s group has also published research that used a Support Vector Machine algorithm to classify cognitively normal individuals from AD patients with 97% accuracy. In addition, a Random Forest classifier was trained to predict conversion from mild cognitive impairment to AD with over 80% accuracy.


The hippocampus is a component of the medial temporal lobe limbic system and plays a central role in the formation, consolidation and retention of recent (or declarative) memory. Atrophy of the hippocampus occurs early in the pathogenesis of AD, which can be detected by structural MRI. Thus, hippocampus atrophy above age expectation has been proposed as a core neuroimaging biomarker of AD. However, measurement of hippocampal atrophy has been challenging in the past. Dr. Ardekani’s group has developed ‘kaiba’ a fully automatic and rapid technique for measuring an index of hippocampal volumetric integrity (HVI) from 3D T1-weighted MRI scans. The group has shown that the bilateral HVI and its time rate of change can be used to reliable differentiate between normal aging, mild cognitive impairment and AD.