| 1. |
- Löfhede, Johan, et al.
(författare)
-
Comparing a Supervised and an Unsupervised Classification Method for Burst Detection in Neonatal EEG
- 2008
-
Ingår i: Proceedings of Engineering in Medicine and Biology Society, EMBS 2008. 30th Annual International Conference of the IEEE, 20-24 August, 2008. - : IEEE. - 1557-170X. - 9781424418145 ; , s. 3836-3839
-
Konferensbidrag (refereegranskat)abstract
- Hidden Markov Models (HMM) and Support Vector Machines (SVM) using unsupervised and supervised training, respectively, were compared with respect to their ability to correctly classify burst and suppression in neonatal EEG. Each classifier was fed five feature signals extracted from EEG signals from six full term infants who had suffered from perinatal asphyxia. Visual inspection of the EEG by an experienced electroencephalographer was used as the gold standard when training the SVM, and for evaluating the performance of both methods. The results are presented as receiver operating characteristic (ROC) curves and quantified by the area under the curve (AUC). Our study show that the SVM and the HMM exhibit similar performance, despite their fundamental differences.
|
|
| 2. |
- Althoff, Karin, 1974, et al.
(författare)
-
Combined segmentation and tracking of neural stem-cells
- 2005
-
Ingår i: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). - 1611-3349 .- 0302-9743. ; 3540, s. 282-291
-
Tidskriftsartikel (refereegranskat)abstract
- In this paper we analyze neural stem/progenitor cells in an time-lapse image sequence. By using information about the previous positions of the cells, we are able to make a. better selection of possible cells out of a collection of blob-like objects. As a blob detector we use Laplacian of Gaussian (LoG) filters at multiple scales, and the cell contours of the selected cells are segmented using dynamic programming. After the segmentation process the cells are tracked in the sequence using a. combined nearest-neighbor and correlation matching technique. An evaluation of the system show that 95% of the cells were correctly segmented and tracked between consecutive frames.
|
|
| 3. |
- Degerman, Johan, 1976, et al.
(författare)
-
An automatic system for in vitro cell migration studies
- 2009
-
Ingår i: Journal of Microscopy. - : Wiley. - 0022-2720 .- 1365-2818. ; 233:1, s. 178-191
-
Tidskriftsartikel (refereegranskat)abstract
- This paper describes a system for in vitro cell migration analysis. Adult neural stem/progenitor cells are studied using time-lapse bright-field microscopy and thereafter stained immunohistochemically to find and distinguish undifferentiated glial progenitor cells and cells having differentiated into type-1 or type-2 astrocytes. The cells are automatically segmented and tracked through the time-lapse sequence. An extension to the Chan-Vese Level Set segmentation algorithm, including two new terms for specialized growing and pruning, made it possible to resolve clustered cells, and reduced the tracking error by 65%. We used a custom-built manual correction module to form a ground truth used as a reference for tracked cells that could be identified from the fluorescence staining. On average, the tracks were correct 95% of the time, using our new segmentation. The tracking, or association of segmented cells, was performed using a 2-state Hidden Markov Model describing the random behaviour of the cells. By re-estimating the motion model to conform with the segmented data we managed to reduce the number of tracking parameters to essentially only one. Upon characterization of the cell migration by the HMM state occupation function, it was found that glial progenitor cells were moving randomly 2/3 of the time, while the type-2 astrocytes showed a directed movement 2/3 of the time. This finding indicates possibilities for cell-type specific identification and cell sorting of live cells based on specific movement patterns in individual cell populations, which would have valuable applications in neurobiological research.
|
|
| 4. |
- Degerman, Johan, 1976, et al.
(författare)
-
Extended Target Tracking using Principal Components
- 2011
-
Ingår i: Proceedings of the 14th International Conference on Information Fusion. - 9781457702679 ; , s. Art. no. 5977659-
-
Konferensbidrag (refereegranskat)abstract
- The increased resolution in today’s radar systemsenables tracking of small targets. However, tracking both smalland large targets in a dense target scenario raises considerablechallenges. The data association of tracks to measurement groupsis highly dependent on good target extension models for filteringand likelihood computation. In our attempt to design a trackerfor extended targets, we start by adopting the results from thetechnique referred to as random matrices, which enables us toseparate the filtering into an extension and a kinematical part.We re-define the measurement model and discard the assumptionof independent Gaussian-distributed plots. Instead we assume theprincipal components to be Gaussian distributed. Then, througha heuristic approach, we create a two-stage Kalman filter, wherethe first stage estimates the principal components, and the secondstage estimates the centre of gravity, using the output fromthe first stage as measurement uncertainty. The advantage ofhaving a Kalman filter with data-driven measurement noise overa standard Kalman filter is demonstrated using simulated data,where a significant improvement in terms of smaller errors andreduced track loss is shown.
|
|
| 5. |
- Degerman, Johan, 1976
(författare)
-
Time-Lapse Bright-Field Microscopy and Image Acquisition of In-Vitro Neural Stem Cells.
- 2005
-
Licentiatavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This project involves development of a system for time-lapse image acquisition of neural stem cells. The purpose of analyzing time sequences is to find migrational patterns that are characteristic for different cell types. Neural stem/progenitor cells posses the capability to differentiate into three lineages: neuronal, astrocytic, and oligodendrocytic. Preliminary results show that glial progenitors migrate significantly more random, than for example astrocytes, that express a more directed movement. Automated image acquisition constitutes the first part of a migration analysis system, followed by an image segmentation and tracking module. It is crucial for the analysis that images are properly acquired. Translation between subsequent images has to be compensated for, and background non-uniformity eliminated. Most important is that the focus level is selected optimally with good repeatability. Optimal in the sense of accurately locating cells, is a criteria difficult to formalize, especially since the bright-field microscope does not produce sufficient contrast for imaging cells in-focus. More contrast was bought at the cost of lower resolution by defocusing, and to find the optimal balance we moved on to work with cell segmentation methods. Inspired by scale-space models of retinal receptive field sampling in human vision, we used multi-scale Laplace of Gaussian with automatic scale selection as a detection filter. The cell shape was computed by tracing the border using dynamic programming. This method show slightly better performance than the watershed segmentation and the multi-scale approach is a promising step towards finding optimal prerequisites for cell location.
|
|
| 6. |
- Degerman, Johan, 1976
(författare)
-
Time-Lapse Microscopy Imaging and Image Analysis for Cell Migration Studies
- 2008
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- This thesis considers a complete automatic system for in-vitro cell imaging and migration image analysis. The development of such can be divided into three main problems:The first considered problem concerns how to build a robust time-lapse sequence acquisition system, for imaging adult hippocampal cells in bright-field microscopy. Due to low contrast the focusing problem becomes ill-defined, and image processing techniques, such as wavelet image fusion, is suggested for obtaining images with high contrast as well as resolution.The second (and main) considered problem is cell image segmentation, i.e how to automatically locate and delineate cells in microscopic images using digital techniques. Cells cultured on glass plates have a tendency to congregate, which makes this task difficult. We study several different methods before arriving at our preferred choice; a modified version of the Chan-Vese level set segmentation. To improve cell segmentation performance, we apply an adaptive multi-scale Laplacian of Gaussian filter. Then, to solve the problem of undetected cells and merged cells (several cells detected as one), we add a new extension to the level set method that we refer to as specialized growing and pruning.The third problem concerns the creation of a motion model for the cells, to aid the tracking and data association and to classify cells on the basis of their migration pattern. We take on several different approaches before concluding that cell migration can be modelled by a non-stationary two-state hidden Markov model. Using this model, we find that glial progenitor cells are moving randomly 2/3 of the time, while type-2 astrocytes show a directed movement 2/3 of the time. This finding indicates possibilities for cell-type specific identification and cell sorting of live migrating cells, which will have a valuable application in neurobiological research.
|
|
| 7. |
- Jonsson, Roland, et al.
(författare)
-
Multi-Target tracking with background discrimination using PHD filters
- 2012
-
Ingår i: Proceedings of the 15th International Conference on Information Fusion.
-
Konferensbidrag (refereegranskat)abstract
- In this paper, we propose a new double PHD filter for simultaneous multi-target tracking and background discrim- ination for airborne radar applications. Both the foreground and the background processes are modeled as Poisson point processes, which gives a symmetric formulation of the coupled filters. The differences between foreground and background lie in the assumed target dynamics, and in the sensor detection probabilities. Although there are proposals for PHD filter with adaptive background models in the literature, our filter appears to be novel and also the simplest possible. To implement the filter we use a Gaussian mixture approximation of the intensities, which enables simple and effective ways to extract tracks. For the evaluations we use a simulated target tracking scenario with an airborne radar tracking a number of flying targets over a background of road objects. First, the performance of the Gaussian mixture PHD filter with track extraction is illustrated. Second, the superior ability of the foreground-background PHD filter to suppress clutter and disturbing road traffic is illustrated.
|
|
| 8. |
|
|
