| 1. |
- Alipoor, Mohammad, 1983, et al.
(author)
-
Determinant of the information matrix: a new rotation invariant optimality metric to design gradient encoding schemes
- 2015
-
In: 12th IEEE International Symposium on Biomedical Imaging, ISBI 2015, Brooklyn, United States, 16-19 April 2015. - 1945-8452. - 9781479923748 ; 2015-July, s. 462-465
-
Conference paper (peer-reviewed)abstract
- Minimum condition number (CN) gradient encoding schemewas introduced to diffusion MRI community more than adecade ago. It’s computation requires tedious numerical optimization which usually leads to sub-optimal solutions. TheCN does not reflect any benefits in acquiring more measurements, i.e. it’s optimal value is constant for any numberof measurements. Further, it is variable under rotation. Inthis paper we (i) propose an accurate method to computeminimum condition number scheme; and (ii) introduce determinant of the information matrix (DIM) as a new optimality metric that scales with number of measurements anddoes reflect what one would gain from acquiring more measurements. Theoretical analysis shows that DIM is rotationinvariant. Evaluations on state-of-the-art encoding schemesproves the relevance and superiority of the proposed metriccompared to condition number.
|
|
| 2. |
- Alipoor, Mohammad, 1983, et al.
(author)
-
Icosahedral gradient encoding scheme for an arbitrary number of measurements
- 2015
-
In: International symposium on biomedical imaging. - 1945-8452. - 9781479923748 ; 2015-July, s. 959-962
-
Conference paper (peer-reviewed)abstract
- The icosahedral gradient encoding scheme (GES) is widelyused in diffusion MRI community due to its uniformly distributed orientations and rotationally invariant condition number. The major drawback with this scheme is that it is notavailable for arbitrary number of measurements. In this paper(i) we propose an algorithm to find the icosahedral schemefor any number of measurements. Performance of the obtained GES is evaluated and compared with that of Jones andtraditional icosahedral schemes in terms of condition number,standard deviation of the estimated fractional anisotropy anddistribution of diffusion sensitizing directions; and (ii) we introduce minimum eigenvalue of the information matrix as anew optimality metric to replace condition number. Unlikecondition number, it is proportional to the number of measurements and thus in agreement with the intuition that moremeasurements leads to more robust tensor estimation. Furthermore, it may independently be maximized to design GESsfor different diffusion imaging techniques.
|
|
| 3. |
- Källén, Hanna, et al.
(author)
-
Towards Grading Gleason Score using Generically Trained Deep convolutional Neural Networks
- 2016
-
In: 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI). - : Institute of Electrical and Electronics Engineers (IEEE). - 9781479923496 - 9781479923502 ; 2016-June, s. 1163-1167
-
Conference paper (peer-reviewed)abstract
- We developed an automatic algorithm with the purpose to assist pathologists to report Gleason score on malignant prostatic adenocarcinoma specimen. In order to detect and classify the cancerous tissue, a deep convolutional neural network that had been pre-trained on a large set of photographic images was used. A specific aim was to support intuitive interaction with the result, to let pathologists adjust and correct the output. Therefore, we have designed an algorithm that makes a spatial classification of the whole slide into the same growth patterns as pathologists do. The 22-layer network was cut at an earlier layer and the output from that layer was used to train both a random forest classifier and a support vector machines classifier. At a specific layer a small patch of the image was used to calculate a feature vector and an image is represented by a number of those vectors. We have classified both the individual patches and the entire images. The classification results were compared for different scales of the images and feature vectors from two different layers from the network. Testing was made on a dataset consisting of 213 images, all containing a single class, benign tissue or Gleason score 3-5. Using 10-fold cross validation the accuracy per patch was 81 %. For whole images, the accuracy was increased to 89 %.
|
|
