| 1. |
- Andersson, Axel, et al.
(författare)
-
End-to-end Multiple Instance Learning with Gradient Accumulation
- 2022
-
Ingår i: 2022 IEEE International Conference on Big Data (Big Data). - : Institute of Electrical and Electronics Engineers (IEEE). - 9781665480451 - 9781665480468 ; , s. 2742-2746
-
Konferensbidrag (refereegranskat)abstract
- Being able to learn on weakly labeled data and provide interpretability are two of the main reasons why attention-based deep multiple instance learning (ABMIL) methods have become particularly popular for classification of histopathological images. Such image data usually come in the form of gigapixel-sized whole-slide-images (WSI) that are cropped into smaller patches (instances). However, the sheer volume of the data poses a practical big data challenge: All the instances from one WSI cannot fit the GPU memory of conventional deep-learning models. Existing solutions compromise training by relying on pre-trained models, strategic selection of instances, sub-sampling, or self-supervised pre-training. We propose a training strategy based on gradient accumulation that enables direct end-to-end training of ABMIL models without being limited by GPU memory. We conduct experiments on both QMNIST and Imagenette to investigate the performance and training time and compare with the conventional memory-expensive baseline as well as a recent sampled-based approach. This memory-efficient approach, although slower, reaches performance indistinguishable from the memory-expensive baseline.
|
|
| 2. |
|
|
| 3. |
|
|
| 4. |
|
|
| 5. |
- Bajic, Buda, et al.
(författare)
-
Blind restoration of images degraded with mixed poisson-Gaussian noise with application in transmission electron microscopy
- 2016
-
Ingår i: 2016 Ieee 13Th International Symposium On Biomedical Imaging (ISBI). - : IEEE. - 9781479923496 - 9781479923502 ; , s. 123-127
-
Konferensbidrag (refereegranskat)abstract
- Noise and blur, present in images after acquisition, negatively affect their further analysis. For image enhancement when the Point Spread Function (PSF) is unknown, blind deblurring is suitable, where both the PSF and the original image are simultaneously reconstructed. In many realistic imaging conditions, noise is modelled as a mixture of Poisson (signal-dependent) and Gaussian (signal independent) noise. In this paper we propose a blind deconvolution method for images degraded by such mixed noise. The method is based on regularized energy minimization. We evaluate its performance on synthetic images, for different blur kernels and different levels of noise, and compare with non-blind restoration. We illustrate the performance of the method on Transmission Electron Microscopy images of cilia, used in clinical practice for diagnosis of a particular type of genetic disorders.
|
|
| 6. |
|
|
| 7. |
- Bajic, Buda, et al.
(författare)
-
Single image super-resolution reconstruction in presence of mixed Poisson-Gaussian noise
- 2016
-
Ingår i: 2016 SIXTH INTERNATIONAL CONFERENCE ON IMAGE PROCESSING THEORY, TOOLS AND APPLICATIONS (IPTA). - : IEEE. - 9781467389105
-
Konferensbidrag (refereegranskat)abstract
- Single image super-resolution (SR) reconstructionaims to estimate a noise-free and blur-free high resolution imagefrom a single blurred and noisy lower resolution observation.Most existing SR reconstruction methods assume that noise in theimage is white Gaussian. Noise resulting from photon countingdevices, as commonly used in image acquisition, is, however,better modelled with a mixed Poisson-Gaussian distribution. Inthis study we propose a single image SR reconstruction methodbased on energy minimization for images degraded by mixedPoisson-Gaussian noise.We evaluate performance of the proposedmethod on synthetic images, for different levels of blur andnoise, and compare it with recent methods for non-Gaussiannoise. Analysis shows that the appropriate treatment of signaldependentnoise, provided by our proposed method, leads tosignificant improvement in reconstruction performance.
|
|
| 8. |
|
|
| 9. |
- Bengtsson Bernander, Karl, et al.
(författare)
-
Rotation-Equivariant Semantic Instance Segmentation on Biomedical Images
- 2022
-
Ingår i: Medical Image Understanding and Analysis, MIUA 2022. - Cham : Springer. - 9783031120534 - 9783031120527 ; , s. 283-297
-
Konferensbidrag (refereegranskat)abstract
- Advances in image segmentation techniques, brought by convolutional neural network (CNN) architectures like U-Net, show promise for tasks such as automated cancer screening. Recently, these methods have been extended to detect different instances of the same class, which could be used to, for example, characterize individual cells in whole-slide images. Still, the amount of data needed and the number of parameters in the network are substantial. To alleviate these problems, we modify a method of semantic instance segmentation to also enforce equivariance to the p4 symmetry group of 90-degree rotations and translations. We perform four experiments on a synthetic dataset of scattered sticks and a subset of the Kaggle 2018 Data Science Bowl, the BBBC038 dataset, consisting of segmented nuclei images. Results indicate that the rotation-equivariant architecture yields similar accuracy as a baseline architecture. Furthermore, we observe that the rotation-equivariant architecture converges faster than the baseline. This is a promising step towards reducing the training time during development of methods based on deep learning.
|
|
| 10. |
- Bengtsson, Ewert, 1948-, et al.
(författare)
-
Detection of Malignancy-Associated Changes Due to Precancerous and Oral Cancer Lesions: A Pilot Study Using Deep Learning
- 2018
-
Ingår i: CYTO2018.
-
Konferensbidrag (refereegranskat)abstract
- Background: The incidence of oral cancer is increasing and it is effecting younger individuals. PAP smear-based screening, visual, and automated, have been used for decades, to successfully decrease the incidence of cervical cancer. Can similar methods be used for oral cancer screening? We have carried out a pilot study using neural networks for classifying cells, both from cervical cancer and oral cancer patients. The results which were reported from a technical point of view at the 2017 IEEE International Conference on Computer Vision Workshop (ICCVW), were particularly interesting for the oral cancer cases, and we are currently collecting and analyzing samples from more patients. Methods: Samples were collected with a brush in the oral cavity and smeared on glass slides, stained, and prepared, according to standard PAP procedures. Images from the slides were digitized with a 0.35 micron pixel size, using focus stacks with 15 levels 0.4 micron apart. Between 245 and 2,123 cell nuclei were manually selected for analysis for each of 14 datasets, usually 2 datasets for each of the 6 cases, in total around 15,000 cells. A small region was cropped around each nucleus, and the best 2 adjacent focus layers in each direction were automatically found, thus creating images of 100x100x5 pixels. Nuclei were chosen with an aim to select well preserved free-lying cells, with no effort to specifically select diagnostic cells. We therefore had no ground truth on the cellular level, only on the patient level. Subsets of these images were used for training 2 sets of neural networks, created according to the ResNet and VGG architectures described in literature, to distinguish between cells from healthy persons, and those with precancerous lesions. The datasets were augmented through mirroring and 90 degrees rotations. The resulting networks were used to classify subsets of cells from different persons, than those in the training sets. This was repeated for a total of 5 folds. Results: The results were expressed as the percentage of cell nuclei that the neural networks indicated as positive. The percentage of positive cells from healthy persons was in the range 8% to 38%. The percentage of positive cells collected near the lesions was in the range 31% to 96%. The percentages from the healthy side of the oral cavity of patients with lesions ranged 37% to 89%. For each fold, it was possible to find a threshold for the number of positive cells that would correctly classify all patients as normal or positive, even for the samples taken from the healthy side of the oral cavity. The network based on the ResNet architecture showed slightly better performance than the VGG-based one. Conclusion: Our small pilot study indicates that malignancyassociated changes that can be detected by neural networks may exist among cells in the oral cavity of patients with precancerous lesions. We are currently collecting samples from more patients, and will present those results as well, with our poster at CYTO 2018.
|
|
