Ahmed Al-Taie

Special Section on Uncertainty and Parameter Space Analysis in Visualization: Uncertainty estimation and visualization in probabilistic segmentation

Probabilistic segmentation is concerned with handling imperfections of image segmentation algorithms. They assign to each voxel and each segment a probability that the voxel belongs to the segment. This is often the starting point for estimating and visualizing uncertainties in the segmentation result. We propose a novel, generally applicable uncertainty estimation approach that considers all probabilities to compute a single uncertainty value between 0 and 1 for each voxel. It is based on aspects of information theory and uses the Kullback-Leibler divergence (or the total variation divergence). We developed several forms of the proposed approach and analyze and compare their behaviors. We show the advantage over existing approaches, derive aggregated uncertainty measures that are useful for judging the accuracy of a probabilistic segmentation algorithm, and present visualization methods to highlight uncertainties in segmentation results.

Uncertainty-aware ensemble of classifiers for segmenting brain MRI data

Estimating and visualizing uncertainty in medical image segmentation has become an active research area due to the necessity of making medical experts aware of possibly wrong segmentation decisions. Still, to our knowledge all these methods are based on a single choice of the underlying segmentation approach. Segmentation using an ensemble of classifiers (or committee machine) use multiple classifiers to increase the performance when compared to applying a single classifier. In this paper, we propose methods to estimate uncertainties in segmentations produced by ensembles of classifiers. We investigate and compare the different combining strategies of the segmentation results of the ensemble members from an uncertainty point of view. We discuss why some combining strategies tend to perform better than others. Also, we visualize the estimated uncertainties using a color mapping in image space and propose a post-segmentations correction step to reclassify the noisy pixels in the final result based on the statistical uncertainty.

Uncertainty estimation and visualization for multi-modal image segmentation

Multi-modal imaging allows for the integration of complementary information from multiple medical imaging modalities for an improved analysis. The multiple information channels may lead to a reduction of the uncertainty in the analysis and decision-making process. Recently, efforts have been made to estimate the uncertainty in uni-modal image segmentation decisions and visually convey this information to the medical experts that examine the image segmentation results. We propose an approach to extend uncertainty estimation and visualization methods to multi-modal image segmentations. We combine probabilistic uni-modal image segmentation results using the concept of ensemble of classifiers. The uncertainty is computed using a measure that is based on the Kullback-Leibler divergence. We apply our approach for an improved segmentation of Multiple Sclerosis (MS) lesions from multiple MR brain imaging modalities. Moreover, we demonstrate how our approach can be used to estimate and visualize the growth of a brain tumor area for imaging data taken at multiple points in time. Both the MS lesion and the area of tumor growth are detected as areas of high uncertainty due to different characteristics in different imaging modalities and changes over time, respectively.

Erratum to: “Combining Rules Using Local Statistics and Uncertainty Estimates for Improved Ensemble Segmentation”

The following affiliation, denoted “d,” should be included for the first author A. Al-Taie: dComputer Science Department, College of Science for Women, Baghdad University, Iraq

Fast Uncertainty-Guided Fuzzy C-Means Segmentation of Medical Images

Image segmentation is a crucial step of the medical visualization pipeline. In this paper, we present a novel fast algorithm for modified fuzzy c-means segmentation of MRI data. The algorithm consists of two steps, which are executed as two iterations of a fuzzy c-means approach: the first iteration is a standard fuzzy c-means (FCM) iteration, while the second iteration is our modified FCM iteration with misclassification correction. In the second iteration, we use the classification probability vectors (uncertainties) of the neighbor pixels found by the first iteration to confirm or correct the classification decision of the current pixel. The application of the proposed algorithm on synthetic data, simulated MRI data, and real MRI data show that our algorithm is insensitive to different types of noise and outperforms the standard FCM and several versions of modified FCM algorithms in terms of accuracy and speed. In fact, our algorithm can easily be combined with many modified FCM algorithms to improve their segmentation result while reducing the computation costs (using two FCM iterations only). An optional simple post-processing step can further improve the segmentation result by correcting isolated misclassified pixels. We also show that our algorithm reduces the uncertainty in the segmentation result, by using recently proposed uncertainty estimation and visualization tools.

Uncertainty and reproducibility in medical visualization

The medical visualization pipeline is affected by various sources of uncertainty. Many errors may occur and several assumptions are made in the various processing steps from the image acquisition to the rendering of the visualization output, which induce uncertainty. High uncertainty leads to low robustness of the algorithms impacting reproducibility of the results. We present how uncertainty can be mathematically described in the medical context. Moreover, in medical applications, the visualization is typically based on a segmentation of the medical images. We propose a method to capture uncertainty in image segmentation and present extensions to ensemble and multi-modal image segmentation.

Point-wise Diversity Measure and Visualization for Ensemble of Classifiers - With Application to Image Segmentation

The idea of using ensembles of classifiers is to increase the performance when compared to applying a single classifier. Crucial to the performance improvement is the diversity of the ensemble. A classifier ensemble is considered to be diverse, if the classifiers make no coinciding errors. Several studies discuss the diversity issue and its relation to the ensemble accuracy. Most of them proposed measures that are based on an ”Oracle” classification. In this paper, we propose a new probability-based diversity measure for ensembles of unsupervised classifiers, i.e., when no Oracle machine exists. Our measure uses a point-wise definition of diversity, which allows for a distinction of diverse and non-diverse areas. Moreover, we introduce the concept of further categorizing the diverse areas into healthy and unhealthy diversity areas. A diversity area is healthy for the ensemble performance, if there is enough redundancy to compensate for the errors. Then, the performance of the ensemble can be based on two parameters, the non-diversity area, i.e., the size of all regions where the classifiers of the ensemble agree, and the healthy diversity area, i.e., the size of the regions where the diversity is healthy. Furthermore, our point-wise diversity measure allows for an intuitive visualization of the ensemble diversity for visual ensemble performance comparison in the context of image segmentation.