CVSnet: A machine learning approach for automated central vein sign assessment in multiple sclerosis

The central vein sign (CVS) is an efficient imaging biomarker for multiple sclerosis (MS) diagnosis, but its application in clinical routine is limited by inter-rater variability and the expenditure of time associated with manual assessment. We describe a deep learning-based prototype for automated assessment of the CVS in white matter MS lesions using data from three different imaging centers. We retrospectively analyzed data from 3 T magnetic resonance images acquired on four scanners from two different vendors, including adults with MS (n = 42), MS mimics (n = 33, encompassing 12 distinct neurological diseases mimicking MS) and uncertain diagnosis (n = 5). Brain white matter lesions were manually segmented on FLAIR* images. Perivenular assessment was performed according to consensus guidelines and used as ground truth, yielding 539 CVS-positive (CVS+) and 448 CVS-negative (CVS-) lesions. A 3D convolutional neural network ("CVSnet") was designed and trained on 47 datasets, keeping 33 for testing. FLAIR* lesion patches of CVS+/CVS- lesions were used for training and validation (n = 375/298) and for testing (n = 164/150). Performance was evaluated lesion-wise and subject-wise and compared with a state-of-the-art vesselness filtering approach through McNemar's test. The proposed CVSnet approached human performance, with lesion-wise median balanced accuracy of 81%, and subject-wise balanced accuracy of 89% on the validation set, and 91% on the test set. The process of CVS assessment, in previously manually segmented lesions, was similar to 600-fold faster using the proposed CVSnet compared with human visual assessment (test set: 4 seconds vs. 40 minutes). On the validation and test sets, the lesion-wise performance outperformed the vesselness filter method (P < 0.001). The proposed deep learning prototype shows promising performance in differentiating MS from its mimics. Our approach was evaluated using data from different hospitals, enabling larger multicenter trials to evaluate the benefit of introducing the CVS marker into MS diagnostic criteria.

Prediction of Survival and Risk Assessment Using Joint Analysis of Microarray Gene Expression Data

Haleh Yasrebi

Gene expression profiles have been widely used in molecular classification, diagnosis and prediction, particularly in the area of oncology where accurate and early diagnosis is needed for appropriate treatment. Avoiding under-/over-treatment when it is not necessary can extend a patient's survival and prevent disease recurrence. These high-throughput assay technologies have generated terabytes of data exploited extensively to provide insights on cancer biology and the underlying mechanism of disease progression. The ultimate goal is to identify possibly tailored treatment and therapy for personalized medicine. Analysis of microarray data is constrained by the following characteristics: (i) noisy due to missing or erroneous values; (ii) high dimensional due to a large number of genes versus a few number of samples in which their expression levels are measured; (iii) costly due to expensive microarray experiments. Abundant microarray gene expression data should be processed by appropriate computational and statistical learning methodologies such as machine learning techniques. These methods are robust to noisy data and have a great capacity to analyze high dimensional data. Their computational power is nevertheless limited to sample size based on which these methods are built. These algorithms have been widely applied to microarray gene expression data to identify a set of genes known as a gene signature whose expressions are highly correlated to a target value or outcome such as disease status, tumor subtype, a patient's survival time, risk of mortality or cancer relapse. Prediction of survival time and a patient's risk which is unknown at diagnosis presents a more challenging task for machine learning methods than tumor subtype or disease classification, which is already established by oncologists. The properties of microarray data cited above, the limitation of the number of samples in cancer patients and dependency of the machine learning methods' performance on sample size justify joint analysis of microarray data to increase the number of samples. We applied joint analysis methods to breast and lung cancer data sets to improve survival prediction and risk assessment. In overall, no significant improvement or deterioration of the performance accuracy was obtained with joint analysis. However, increasing sample size helped to identify robust or stable gene signatures predictive of survival time and risk assessment. Our achievements and learned-lessons from joint analysis of microarray gene expression data can be used as a guideline for future research studies in classification and prediction.

EPFL2010

Learning Discriminative Features and Structured Models for Segmentation in Microscopy and Natural Images

Aurélien Lucchi

Segmenting images is a significant challenge that has drawn a lot of attention from different fields of artificial intelligence and has many practical applications. One such challenge addressed in this thesis is the segmentation of electron microscope (EM) imaging of neural tissue. EM microscopy is one of the key tools used to analyze neural tissue and understand the brain, but the huge amounts of data it produces make automated analysis necessary. In addition to the challenges specific to EM data, the common problems encountered in image segmentation must also be addressed. These problems include extracting discriminative features from the data and constructing a statistical model using ground-truth data. Although complex models appear to be more attractive because they allow for more expressiveness, they also lead to a higher computational complexity. On the other hand, simple models come with a lower complexity but less faithfully express the real world. Therefore, one of the most challenging tasks in image segmentation is in constructing models that are expressive enough while remaining tractable. In this work, we propose several automated graph partitioning approaches that address these issues. These methods reduce the computational complexity by operating on supervoxels instead of voxels, incorporating features capable of describing the 3D shape of the target objects and using structured models to account for correlation in output variables. One of the non-trivial issues with such models is that their parameters must be carefully chosen for optimal performance. A popular approach to learning model parameters is a maximum-margin approach called Structured SVM (SSVM) that provides optimality guarantees but also suffers from two main drawbacks. First, SSVM-based approaches are usually limited to linear kernels, since more powerful nonlinear kernels cause the learning to become prohibitively expensive. In this thesis, we introduce an approach to “kernelize” the features so that a linear SSVM framework can leverage the power of nonlinear kernels without incurring their high computational cost. Second, the optimality guarentees are violated for complex models with strong inter-relations between the output variables. We propose a new subgradient-based method that is more robust and leads to improved convergence properties and increased reliability. The different approaches presented in this thesis are applicable to both natural and medical images. They are able to segment mitochondria at a performance level close to that of a human annotator, and outperform state-of-the-art segmentation techniques while still benefiting from a low learning time.

EPFL2013

Learning Discriminative Features and Structured Models for Segmentation in Microscopy and Natural Images

Aurélien Lucchi

EPFL2013

Prediction of Survival and Risk Assessment Using Joint Analysis of Microarray Gene Expression Data

Haleh Yasrebi

EPFL2010