Several attempts have been made in the past few years to develop and implement an automated segmentation of neonatal brain structural MRI. However, accurate automated MRI segmentation remains challenging in this population because of the low signal-to-noise ratio, large partial volume effects and inter-individual anatomical variability of the neonatal brain. In this paper, we propose a learning method for segmenting the whole brain cortical grey matter on neonatal T2-weighted images. We trained our algorithm using a neonatal dataset composed of 3 fullterm and 4 preterm infants scanned at term equivalent age. Our segmentation pipeline combines the FAST algorithm from the FSL library software and a Bayesian segmentation approach to create a threshold matrix that minimizes the error of mislabeling brain tissue types. Our method shows promising results with our pilot training set. In both preterm and full-term neonates, automated Bayesian segmentation generates a smoother and more consistent parcellation compared to FAST, while successfully removing the subcortical structure and cleaning the edges of the cortical grey matter. This method show promising refinement of the FAST segmentation by considerably reducing manual input and editing required from the user, and further improving reliability and processing time of neonatal MR images. Further improvement will include a larger dataset of training images acquired from different manufacturers.
Lower back pain is one of the most prevalent disorders in the developed/developing world. However, its etiology is poorly understood and treatment is often determined subjectively. In order to quantitatively study the emergence and evolution of back pain, it is necessary to develop consistently measurable markers for pathology. Imaging based measures offer one solution to this problem. The development of imaging based on quantitative biomarkers for the lower back necessitates automated techniques to acquire this data. While the problem of segmenting lumbar vertebrae has been addressed repeatedly in literature, the associated problem of computing relevant biomarkers on the basis of the segmentation has not been addressed thoroughly. In this paper, we propose a Random-Forest based approach that learns to segment vertebral bodies in CT images followed by a biomarker evaluation framework that extracts vertebral heights and widths from the segmentations obtained. Our dataset consists of 15 CT sagittal scans obtained from General Electric Healthcare. Our main approach is divided into three parts: the first stage is image pre-processing which is used to correct for variations in illumination across all the images followed by preparing the foreground and background objects from images; the next stage is Machine Learning using Random-Forests, which distinguishes the interest-point vectors between foreground or background; and the last step is image post-processing, which is crucial to refine the results of classifier. The Dice coefficient was used as a statistical validation metric to evaluate the performance of our segmentations with an average value of 0.725 for our dataset.
Lower back pain and pathologies related to it are one of the most common results for a referral to a neurosurgical clinic in the developed and the developing world. Quantitative evaluation of these pathologies is a challenge. Image based measurements of angles/vertebral heights and disks could provide a potential quantitative biomarker for tracking and measuring these pathologies. Detection of vertebral bodies is a key element and is the focus of the current work. From the variety of medical imaging techniques, MRI and CT scans have been typically used for developing image segmentation methods. However, CT scans are known to give a large dose of x-rays, increasing cancer risk . MRI can be substituted for CTs when the risk is high  but are difficult to obtain in smaller facilities due to cost and lack of expertise in the field . X-rays provide another option with its ability to control the x-ray dosage, especially for young people, and its accessibility for smaller facilities. Hence, the ability to create quantitative biomarkers from x-ray data is especially valuable. Here, we develop a multiscale template matching, inspired by , to detect centers of vertebral bodies from x-ray data. The immediate application of such detection lies in developing quantitative biomarkers and in querying similar images in a database. Previously, shape similarity classification methods have been used to address this problem, but these are challenging to use in the presence of variation due to gross pathology and even subtle effects .
Deep Learning, refers to large set of neural network based algorithms, have emerged as promising machine- learning tools in the general imaging and computer vision domains. Convolutional neural networks (CNNs), a specific class of deep learning algorithms, have been extremely effective in object recognition and localization in natural images. A characteristic feature of CNNs, is the use of a locally connected multi layer topology that is inspired by the animal visual cortex (the most powerful vision system in existence). While CNNs, perform admirably in object identification and localization tasks, typically require training on extremely large datasets. Unfortunately, in medical image analysis, large datasets are either unavailable or are extremely expensive to obtain. Further, the primary tasks in medical imaging are organ identification and segmentation from 3D scans, which are different from the standard computer vision tasks of object recognition. Thus, in order to translate the advantages of deep learning to medical image analysis, there is a need to develop deep network topologies and training methodologies, that are geared towards medical imaging related tasks and can work in a setting where dataset sizes are relatively small. In this paper, we present a technique for stacked supervised training of deep feed forward neural networks for segmenting organs from medical scans. Each `neural network layer' in the stack is trained to identify a sub region of the original image, that contains the organ of interest. By layering several such stacks together a very deep neural network is constructed. Such a network can be used to identify extremely small regions of interest in extremely large images, inspite of a lack of clear contrast in the signal or easily identifiable shape characteristics. What is even more intriguing is that the network stack achieves accurate segmentation even when it is trained on a single image with manually labelled ground truth. We validate this approach,using a publicly available head and neck CT dataset. We also show that a deep neural network of similar depth, if trained directly using backpropagation, cannot acheive the tasks achieved using our layer wise training paradigm.
Accurate and efficient automated tumor segmentation in breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is highly desirable for computer-aided tumor diagnosis. We propose a novel automatic segmentation framework which incorporates mean-shift smoothing, superpixel-wise classification, pixel-wise graph-cuts partitioning, and morphological refinement. A set of 15 breast DCE-MR images, obtained from the American College of Radiology Imaging Network (ACRIN) 6657 I-SPY trial, were manually segmented to generate tumor masks (as ground truth) and breast masks (as regions of interest). Four state-of-the-art segmentation approaches based on diverse models were also utilized for comparison. Based on five standard evaluation metrics for segmentation, the proposed framework consistently outperformed all other approaches. The performance of the proposed framework was: 1) 0.83 for Dice similarity coefficient, 2) 0.96 for pixel-wise accuracy, 3) 0.72 for VOC score, 4) 0.79 mm for mean absolute difference, and 5) 11.71 mm for maximum Hausdorff distance, which surpassed the second best method (i.e., adaptive geodesic transformation), a semi-automatic algorithm depending on precise initialization. Our results suggest promising potential applications of our segmentation framework in assisting analysis of breast carcinomas.
Vertebral segmentation is a critical first step in any quantitative evaluation of vertebral pathology using CT images. This is especially challenging because bone marrow tissue has the same intensity profile as the muscle surrounding the bone. Thus simple methods such as thresholding or adaptive k-means fail to accurately segment vertebrae. While several other algorithms such as level sets may be used for segmentation any algorithm that is clinically deployable has to work in under a few seconds. To address these dual challenges we present here, a new algorithm based on the geodesic distance transform that is capable of segmenting the spinal vertebrae in under one second. To achieve this we extend the theory of the geodesic distance transforms proposed in1 to incorporate high level anatomical knowledge through adaptive weighting of image gradients. Such knowledge may be provided by the user directly or may be automatically generated by another algorithm. We incorporate information 'learnt' using a previously published machine learning algorithm2 to segment the L1 to L5 vertebrae. While we present a particular application here, the adaptive geodesic transform is a generic concept which can be applied to segmentation of other organs as well.