Fritz Albregtsen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Fritz Albregtsen has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Akbar Akbari
Department of Informatics, University of Oslo, Oslo, Norway
1 shared publication - Andreas Austeng
Department of Informatics, University of Oslo, Oslo, Norway
1 shared publication - Andreas Kleppe
Institute for Cancer Genetics and Informatics, Oslo University Hospital, Oslo, Norway; Department of Informatics, Faculty of Mathematics and Natural Sciences, University of Oslo, Oslo, Norway; Centre for Research-based Innovation Visual Intelligence, UiT The Arctic University of Norway, Tromsø, Norway
1 shared publication - Anne H. Schistad Solberg
Department of Informatics, University of Oslo, Oslo, Norway
1 shared publication - Birgitte Jyding Vennervald
Department of Veterinary and Animal Sciences, University of Copenhagen, DYRLAEGEVEJ100.2, DK-1870 Frederiksberg C, Denmark
1 shared publication - Birgitte Nielsen
Department of Medical Informatics, The Norwegian Radium Hospital, Montebello, N-0310 Oslo, Norway
1 shared publication
Latest Publications
- Open Access
- 10 February 2021
- Published by MDPI AG in Electronics
- Vol. 10 (4) , 431
Numerical Evaluation on Parametric Choices Influencing Segmentation Results in Radiology Images—A Multi-Dataset Study
Medical image segmentation has gained greater attention over the past decade, especially in the field of image-guided surgery. Here, robust, accurate and fast segmentation tools are important for planning and navigation. In this work, we explore the Convolutional Neural Network (CNN) based approaches for multi-dataset segmentation from CT examinations. We hypothesize that selection of certain parameters in the network architecture design critically influence the segmentation results. We have employed two different CNN architectures, 3D-UNet and VGG-16, given that both networks are well accepted in the medical domain for segmentation tasks. In order to understand the efficiency of different parameter choices, we have adopted two different approaches. The first one combines different weight initialization schemes with different activation functions, whereas the second approach combines different weight initialization methods with a set of loss functions and optimizers. For evaluation, the 3D-UNet was trained with the Medical Segmentation Decathlon dataset and VGG-16 using LiTS data. The quality assessment done using eight quantitative metrics enhances the probability of using our proposed strategies for enhancing the segmentation results. Following a systematic approach in the evaluation of the results, we propose a few strategies that can be adopted for obtaining good segmentation results. Both of the architectures used in this work were selected on the basis of general acceptance in segmentation tasks for medical images based on their promising results compared to other state-of-the art networks. The highest Dice score obtained in 3D-UNet for the liver, pancreas and cardiac data was 0.897, 0.691 and 0.892. In the case of VGG-16, it was solely developed to work with liver data and delivered a Dice score of 0.921. From all the experiments conducted, we observed that two of the combinations with Xavier weight initialization (also known as Glorot), Adam optimiser, Cross Entropy loss () and LeCun weight initialization, cross entropy loss and Adam optimiser worked best for most of the metrics in a 3D-UNet setting, while Xavier together with cross entropy loss and Tanh activation function () worked best for the VGG-16 network. Here, the parameter combinations are proposed on the basis of their contributions in obtaining optimal outcomes in segmentation evaluations. Moreover, we discuss that the preliminary evaluation results show that these parameters could later on be used for gaining more insights into model convergence and optimal solutions.The results from the quality assessment metrics and the statistical analysis validate our conclusions and we propose that the presented work can be used as a guide in choosing parameters for the best possible segmentation results for future works.- 10 February 2021
- Published by MDPI AG in Electronics
- Vol. 10 (4) , 431
Open AccessNumerical Evaluation on Parametric Choices Influencing Segmentation Results in Radiology Images—A Multi-Dataset Study
Medical image segmentation has gained greater attention over the past decade, especially in the field of image-guided surgery. Here, robust, accurate and fast segmentation tools are important for planning and navigation. In this work, we explore the Convolutional Neural Network (CNN) based approaches for multi-dataset segmentation from CT examinations. We hypothesize that selection of certain parameters in the network architecture design critically influence the segmentation results. We have employed two different CNN architectures, 3D-UNet and VGG-16, given that both networks are well accepted in the medical domain for segmentation tasks. In order to understand the efficiency of different parameter choices, we have adopted two different approaches. The first one combines different weight initialization schemes with different activation functions, whereas the second approach combines different weight initialization methods with a set of loss functions and optimizers. For evaluation, the 3D-UNet was trained with the Medical Segmentation Decathlon dataset and VGG-16 using LiTS data. The quality assessment done using eight quantitative metrics enhances the probability of using our proposed strategies for enhancing the segmentation results. Following a systematic approach in the evaluation of the results, we propose a few strategies that can be adopted for obtaining good segmentation results. Both of the architectures used in this work were selected on the basis of general acceptance in segmentation tasks for medical images based on their promising results compared to other state-of-the art networks. The highest Dice score obtained in 3D-UNet for the liver, pancreas and cardiac data was 0.897, 0.691 and 0.892. In the case of VGG-16, it was solely developed to work with liver data and delivered a Dice score of 0.921. From all the experiments conducted, we observed that two of the combinations with Xavier weight initialization (also known as Glorot), Adam optimiser, Cross Entropy loss () and LeCun weight initialization, cross entropy loss and Adam optimiser worked best for most of the metrics in a 3D-UNet setting, while Xavier together with cross entropy loss and Tanh activation function () worked best for the VGG-16 network. Here, the parameter combinations are proposed on the basis of their contributions in obtaining optimal outcomes in segmentation evaluations. Moreover, we discuss that the preliminary evaluation results show that these parameters could later on be used for gaining more insights into model convergence and optimal solutions.The results from the quality assessment metrics and the statistical analysis validate our conclusions and we propose that the presented work can be used as a guide in choosing parameters for the best possible segmentation results for future works. - Open Access
Prognostic Value of the Diversity of Nuclear Chromatin Compartments in Gynaecological Carcinomas
Statistical texture analysis of cancer cell nuclei stained for DNA has recently been used to develop a pan-cancer prognostic marker of chromatin heterogeneity. In this study, we instead analysed chromatin organisation by automatically quantifying the diversity of chromatin compartments in cancer cell nuclei. The aim was to investigate the prognostic value of such an assessment in relation to chromatin heterogeneity and as a potential supplement to pathological risk classifications in gynaecological carcinomas. The diversity was quantified by calculating the entropy of both chromatin compartment sizes and optical densities within compartments. We analysed a median of 281 nuclei (interquartile range (IQR), 273 to 289) from 246 ovarian carcinoma patients and a median of 997 nuclei (IQR, 502 to 1452) from 791 endometrial carcinoma patients. The prognostic value of the entropies and chromatin heterogeneity was moderately strongly correlated (r ranged from 0.68 to 0.73), but the novel marker was observed to provide additional prognostic information. In multivariable analysis with clinical and pathological markers, the hazard ratio associated with the novel marker was 2.1 (95% CI, 1.3 to 3.5) in ovarian carcinoma and 2.4 (95% CI, 1.5 to 3.9) in endometrial carcinoma. Integration with pathological risk classifications gave three risk groups with distinctly different prognoses. This suggests that the novel marker of diversity of chromatin compartments might possibly contribute to the selection of high-risk stage I ovarian carcinoma patients for adjuvant chemotherapy and low-risk endometrial carcinoma patients for less extensive surgery.
Open AccessPrognostic Value of the Diversity of Nuclear Chromatin Compartments in Gynaecological Carcinomas
Statistical texture analysis of cancer cell nuclei stained for DNA has recently been used to develop a pan-cancer prognostic marker of chromatin heterogeneity. In this study, we instead analysed chromatin organisation by automatically quantifying the diversity of chromatin compartments in cancer cell nuclei. The aim was to investigate the prognostic value of such an assessment in relation to chromatin heterogeneity and as a potential supplement to pathological risk classifications in gynaecological carcinomas. The diversity was quantified by calculating the entropy of both chromatin compartment sizes and optical densities within compartments. We analysed a median of 281 nuclei (interquartile range (IQR), 273 to 289) from 246 ovarian carcinoma patients and a median of 997 nuclei (IQR, 502 to 1452) from 791 endometrial carcinoma patients. The prognostic value of the entropies and chromatin heterogeneity was moderately strongly correlated (r ranged from 0.68 to 0.73), but the novel marker was observed to provide additional prognostic information. In multivariable analysis with clinical and pathological markers, the hazard ratio associated with the novel marker was 2.1 (95% CI, 1.3 to 3.5) in ovarian carcinoma and 2.4 (95% CI, 1.5 to 3.9) in endometrial carcinoma. Integration with pathological risk classifications gave three risk groups with distinctly different prognoses. This suggests that the novel marker of diversity of chromatin compartments might possibly contribute to the selection of high-risk stage I ovarian carcinoma patients for adjuvant chemotherapy and low-risk endometrial carcinoma patients for less extensive surgery.
- Open Access
- 8 July 2015
- Published by American Society of Tropical Medicine and Hygiene in The American Journal of Tropical Medicine and Hygiene
- Vol. 93 (1) , 80-86
The First Step Toward Diagnosing Female Genital Schistosomiasis by Computer Image Analysis
Schistosoma haematobium causes female genital schistosomiasis (FGS), which is a poverty-related disease in sub-Saharan Africa. Furthermore, it is co-endemic with human immunodeficiency virus (HIV), and biopsies from genital lesions may expose the individual to increased risk of HIV infection. However, microscopy of urine and hematuria are nonspecific and insensitive predictors of FGS and gynecological investigation requires extensive training. Safe and affordable diagnostic methods are needed. We explore a novel method of diagnosing FGS using computer color analysis of colposcopic images. In a cross-sectional study on young women in an endemic area, we found strong associations between the output from the computer color analysis and both clinical diagnosis (odds ratio [OR] = 5.97, P < 0.001) and urine microscopy for schistosomiasis (OR = 3.52, P = 0.004). Finally, using latent class statistics, we estimate that the computer color analysis yields a sensitivity of 80.5% and a specificity of 66.2% for the diagnosis of FGS.
- 8 July 2015
- Published by American Society of Tropical Medicine and Hygiene in The American Journal of Tropical Medicine and Hygiene
- Vol. 93 (1) , 80-86
Open AccessThe First Step Toward Diagnosing Female Genital Schistosomiasis by Computer Image Analysis
Schistosoma haematobium causes female genital schistosomiasis (FGS), which is a poverty-related disease in sub-Saharan Africa. Furthermore, it is co-endemic with human immunodeficiency virus (HIV), and biopsies from genital lesions may expose the individual to increased risk of HIV infection. However, microscopy of urine and hematuria are nonspecific and insensitive predictors of FGS and gynecological investigation requires extensive training. Safe and affordable diagnostic methods are needed. We explore a novel method of diagnosing FGS using computer color analysis of colposcopic images. In a cross-sectional study on young women in an endemic area, we found strong associations between the output from the computer color analysis and both clinical diagnosis (odds ratio [OR] = 5.97, P < 0.001) and urine microscopy for schistosomiasis (OR = 3.52, P = 0.004). Finally, using latent class statistics, we estimate that the computer color analysis yields a sensitivity of 80.5% and a specificity of 66.2% for the diagnosis of FGS.
- 15 June 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 53 (11) , 6252-6273
A Radon-Transform-Based Image Noise Filter—With Applications to Multibeam Bathymetry
This paper describes a linear-image-transform-based algorithm for reducing stripe noise, track line artifacts, and motion-induced errors in remote sensing data. Developed for multibeam bathymetry (MB), the method has also been used for removing scalloping in synthetic aperture radar images. The proposed image transform is the composition of an invertible edge detection operator and a fast discrete Radon transform (DRT) due to Götz, Druckmüller, and Brady. The inverse DRT is computed by using an iterative method and exploiting an approximate inverse algorithm due to Press. The edge operator is implemented by circular convolution with a Laplacian point spread function modified to render the operator invertible. In the transformed image, linear discontinuities appear as high-intensity spots, which may be reset to zero. In MB data, a second noise signature is linked to motion-induced errors. A Chebyshev approximation of the original image is subtracted before applying the transform, and added back to the denoised image; this is necessary to avoid boundary effects. It is possible to process data faster and suppress motion-induced noise further by filtering images in nonoverlapping blocks using a matrix representation for the inverse DRT. Processed test images from several MB data sets had less noise and distortion compared with those obtained with standard low-pass filters. Denoising also improved the accuracy in statistical classification of geomorphological type by 10-28% for two sets of invariant terrain features.
- 15 June 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 53 (11) , 6252-6273
A Radon-Transform-Based Image Noise Filter—With Applications to Multibeam Bathymetry
This paper describes a linear-image-transform-based algorithm for reducing stripe noise, track line artifacts, and motion-induced errors in remote sensing data. Developed for multibeam bathymetry (MB), the method has also been used for removing scalloping in synthetic aperture radar images. The proposed image transform is the composition of an invertible edge detection operator and a fast discrete Radon transform (DRT) due to Götz, Druckmüller, and Brady. The inverse DRT is computed by using an iterative method and exploiting an approximate inverse algorithm due to Press. The edge operator is implemented by circular convolution with a Laplacian point spread function modified to render the operator invertible. In the transformed image, linear discontinuities appear as high-intensity spots, which may be reset to zero. In MB data, a second noise signature is linked to motion-induced errors. A Chebyshev approximation of the original image is subtracted before applying the transform, and added back to the denoised image; this is necessary to avoid boundary effects. It is possible to process data faster and suppress motion-induced noise further by filtering images in nonoverlapping blocks using a matrix representation for the inverse DRT. Processed test images from several MB data sets had less noise and distortion compared with those obtained with standard low-pass filters. Denoising also improved the accuracy in statistical classification of geomorphological type by 10-28% for two sets of invariant terrain features.
- 10 August 2010
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 2 (9) , 1360-1371
Automatic lane detection and separation in one dimensional gel images using continuous wavelet transform
This work presents new methods which automatically detect and separate the lanes in one dimensional electrophoresis gel images. The motivation is to present algorithms which can be applied on one dimensional gel images produced from separation of DNA, RNA, (including PCR products), proteins, etc. In addition, a one dimensional intensity profile for each separated lane analogous to capillary electrophoresis (CE) profiles is provided. Experiments are performed on ten multi- and ten single-locus DNA gel images, where the number of lanes, resolution and image quality are highly variable. The Elder & Southern (ES) semi-automatic lane detection method, followed by two new methods: an iterative moving average filter (IMA) and a continuous wavelet transform (CWT) with Morlet wavelet are presented and applied on the test images. The accuracy of the results obtained by applying these methods on a set of 10 multi-locus gel images was 91.7% with the ES, 94.4% with the IMA and 99.5% with the CWT; in the set of 10 single-locus images it was 100% with the ES, 94.8% with the IMA and 97.8% with the CWT; for both sets combined (20 images) it was 96.1% with the ES, 94.7% with the IMA and 98.7% with the CWT. Two lane separation methods are presented; the first is based on a K-neighbouring points algorithm, while a half-width algorithm is applied in the second method. Every separated lane from a gel image can be transferred into a sub-image and further into a one dimensional intensity profile.
- 10 August 2010
- Published by Royal Society of Chemistry (RSC) in Analytical Methods
- Vol. 2 (9) , 1360-1371
Automatic lane detection and separation in one dimensional gel images using continuous wavelet transform
This work presents new methods which automatically detect and separate the lanes in one dimensional electrophoresis gel images. The motivation is to present algorithms which can be applied on one dimensional gel images produced from separation of DNA, RNA, (including PCR products), proteins, etc. In addition, a one dimensional intensity profile for each separated lane analogous to capillary electrophoresis (CE) profiles is provided. Experiments are performed on ten multi- and ten single-locus DNA gel images, where the number of lanes, resolution and image quality are highly variable. The Elder & Southern (ES) semi-automatic lane detection method, followed by two new methods: an iterative moving average filter (IMA) and a continuous wavelet transform (CWT) with Morlet wavelet are presented and applied on the test images. The accuracy of the results obtained by applying these methods on a set of 10 multi-locus gel images was 91.7% with the ES, 94.4% with the IMA and 99.5% with the CWT; in the set of 10 single-locus images it was 100% with the ES, 94.8% with the IMA and 97.8% with the CWT; for both sets combined (20 images) it was 96.1% with the ES, 94.7% with the IMA and 98.7% with the CWT. Two lane separation methods are presented; the first is based on a K-neighbouring points algorithm, while a half-width algorithm is applied in the second method. Every separated lane from a gel image can be transferred into a sub-image and further into a one dimensional intensity profile.
- 1 January 2002
- Published by Springer Nature
- p. 55-65
Fractal Signature Vectors and Lacunarity Class Distance Matrices to Extract New Adaptive Texture Features from Cell Nuclei
- 1 January 2002
- Published by Springer Nature
- p. 55-65
Fractal Signature Vectors and Lacunarity Class Distance Matrices to Extract New Adaptive Texture Features from Cell Nuclei