| 1. |
|
|
| 2. |
|
|
| 3. |
- Ge, Chenjie, 1991, et al.
(författare)
-
Multiscale Deep Convolutional Networks for Characterization and Detection of Alzheimer's Disease using MR Images
- 2019
-
Ingår i: Proceedings - International Conference on Image Processing, ICIP. - 1522-4880. ; 2019-September, s. 789-793
-
Konferensbidrag (refereegranskat)abstract
- This paper addresses the issues of Alzheimer's disease (AD) characterization and detection from Magnetic Resonance Images (MRIs). Many existing AD detection methods use single-scale feature learning from brain scans. In this paper, we propose a multiscale deep learning architecture for learning AD features. The main contributions of the paper include: (a) propose a novel 3D multiscale CNN architecture for the dedicated task of AD detection; (b) propose a feature fusion and enhancement strategy for multiscale features; (c) empirical study on the impact of several settings, including two dataset partitioning approaches, and the use of multiscale and feature enhancement. Experiments were conducted on an open ADNI dataset (1198 brain scans from 337 subjects), test results have shown the effectiveness of the proposed method with test accuracy of 93.53%, 87.24% (best, average) on subject separated dataset, and 99.44%, 98.80% (best, average) on random brain scan-partitioned dataset. Comparison with eight existing methods has provided further support to the proposed method.
|
|
| 4. |
- Ali, Muhaddisa Barat, 1986, et al.
(författare)
-
Multi-stream Convolutional Autoencoder and 2D Generative Adversarial Network for Glioma Classification
- 2019
-
Ingår i: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). - Cham : Springer International Publishing. - 1611-3349 .- 0302-9743. ; 11678 LNCS, s. 234-245
-
Konferensbidrag (refereegranskat)abstract
- Diagnosis and timely treatment play an important role in preventing brain tumor growth. Deep learning methods have gained much attention lately. Obtaining a large amount of annotated medical data remains a challenging issue. Furthermore, high dimensional features of brain images could lead to over-fitting. In this paper, we address the above issues. Firstly, we propose an architecture for Generative Adversarial Networks to generate good quality synthetic 2D MRIs from multi-modality MRIs (T1 contrast-enhanced, T2, FLAIR). Secondly, we propose a deep learning scheme based on 3-streams of Convolutional Autoencoders (CAEs) followed by sensor information fusion. The rational behind using CAEs is that it may improve glioma classification performance (as comparing with conventional CNNs), since CAEs offer noise robustness and also efficient feature reduction hence possibly reduce the over-fitting. A two-round training strategy is also applied by pre-training on GAN augmented synthetic MRIs followed by refined-training on original MRIs. Experiments on BraTS 2017 dataset have demonstrated the effectiveness of the proposed scheme (test accuracy 92.04%). Comparison with several exiting schemes has provided further support to the proposed scheme.
|
|
| 5. |
- Saboori, Arash, et al.
(författare)
-
Unsupervised segmentation and data augmentation in image sequences of skeletal muscle contraction by cycle-consistent generative adversarial network
- 2023
-
Ingår i: 2023 international conference on modeling, simulation & intelligent computing (MoSICom). - : IEEE. - 9798350393415 - 9798350393422 ; , s. 474-479
-
Konferensbidrag (refereegranskat)abstract
- This paper investigates a method addressing the unsupervised segmentation and joint data augmentation in medical ultrasound imaging based on the modified CycleGAN. Accurate quantification of fascia and muscle is the key for the diagnostics of neuromuscular disorders based on the analysis of image sequences of skeletal muscle contraction. Although the Deep Learning (DL) models represent encouraging results, some challenges exist. The traditional models don't consider the complex interaction between tissues within a muscle and its surroundings, which reduces the performance of the fascia segmentation. Also, the DL requires many annotated datasets, which ignores dealing with noisy and complex ultrasound images. To overcome these issues, we propose a method to generate realistic images, and then present an unsupervised fascia segmentation method. The results show that our method improves the segmentation accuracy in noisy and complex ultrasound images compared to the traditional methods.
|
|
| 6. |
- Liu, Xixi, 1995, et al.
(författare)
-
Deep Nearest Neighbors for Anomaly Detection in Chest X-Rays
- 2024
-
Ingår i: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). - 1611-3349 .- 0302-9743. ; 14349 LNCS, s. 293-302
-
Konferensbidrag (refereegranskat)abstract
- Identifying medically abnormal images is crucial to the diagnosis procedure in medical imaging. Due to the scarcity of annotated abnormal images, most reconstruction-based approaches for anomaly detection are trained only with normal images. At test time, images with large reconstruction errors are declared abnormal. In this work, we propose a novel feature-based method for anomaly detection in chest x-rays in a setting where only normal images are provided during training. The model consists of lightweight adaptor and predictor networks on top of a pre-trained feature extractor. The parameters of the pre-trained feature extractor are frozen, and training only involves fine-tuning the proposed adaptor and predictor layers using Siamese representation learning. During inference, multiple augmentations are applied to the test image, and our proposed anomaly score is simply the geometric mean of the k-nearest neighbor distances between the augmented test image features and the training image features. Our method achieves state-of-the-art results on two challenging benchmark datasets, the RSNA Pneumonia Detection Challenge dataset, and the VinBigData Chest X-ray Abnormalities Detection dataset. Furthermore, we empirically show that our method is robust to different amounts of anomalies among the normal images in the training dataset. The code is available at: https://github.com/XixiLiu95/deep-kNN-anomaly-detection.
|
|
| 7. |
- Björkman, Kristoffer, et al.
(författare)
-
Genotype-phenotype correlations in patients with complex I deficiency due to mutations in NDUFS1 and NDUFV1
- 2014
-
Ingår i: Euromit 2014, 15-19 juni, Tampere, Finland.
-
Konferensbidrag (övrigt vetenskapligt/konstnärligt)abstract
- Objectives: To study genotype-phenotype correlations in genes encoding complex I electron input module subunits. Materials and methods: We studied five patients with isolated complex I deficiency, three with NDUFS1 mutations and two with NDUFV1 mutations. A literature review of all reported cases of mutations in the affected genes was performed. Results: The literature review revealed pathological mutations in NDUFS1 for 18 patients in 17 families and correspondingly in NDUFV1 for 26 patients in 19 families. Unpublished clinical data for our five patients were added. Our study showed quite variable clinical courses; death before two years of age was seen in 41% of patients while 18% were alive at seven years. There was a significant difference between the NDUFS1 and NDUFV1 groups for clinical onset and life-span. Mutations in NDUFS1 were linked to a worse clinical course with earlier onset and earlier death. Conclusions: Genotype-phenotype correlations in patients with mutations affecting the genes that encode the electron input module of complex I vary, but patients with NDUFS1 mutation tend to have a worse clinical course than patients with NDUFV1 mutation. Identifying the mutations is of importance for accurate prognostic information and genetic counseling.
|
|
| 8. |
- Mamontov, Eugen, 1955, et al.
(författare)
-
Oncogenic hyperplasia caused by combination of various factors: A decision-support software for radionuclide therapy
- 2007
-
Ingår i: Workshop "Mathematical Modelling and Analysis of Cancer Invasion of Tissues", Mar 26, 2007 - Mar 30, 2007, Dundee, Scotland.
-
Konferensbidrag (refereegranskat)abstract
- The present work deals with the software based on the PhasTraM model [1] for oncogenic hyperplasia, the first stage of formation of any solid tumor. The work generalizes the related results of [2]-[6] and discusses application of the software for decision support in radionuclide therapy. The software capabilities to allow for combinations of various causes of oncogeny are emphasized. The causes comprise inflammation, immune dysfunction, and chronic psychological stress. The immune dysfunction is represented with hypogammaglobulenimia expressed in terms of the concentration of the immunoglobulin-G molecules. The level of chronic pychological stress is described with the concentration of the interleukin-6 molecules. The work considers how application of the software can support decisions on the specific radionuclide-therapy setting depending on the tissue-, organ-, and patient-specific data. This is illustrated by a number of numerical-simulation results, also the ones which include the effects of common and fractionation-based radionuclide-therapy modalities. A proper attention is paid to how specifically the input data can be prepared by prospective users of the software, i.e. the specialists who apply radionuclide therapy. The work also formulates a few directions for future research in connection with the features of the everyday work of the prospective users. REFERENCES: [1] E. Mamontov, K. Psiuk-Maksymowicz, A. Koptioug, 2006, Stochastic mechanics in the context of the properties of living systems, Mathl Comput. Modelling, 44(7-8) 595-607. [2] E. Mamontov, A. V. Koptioug, K. Psiuk-Maksymowicz, 2006, The minimal, phase-transition model for the cell-number maintenance by the hyperplasia-extended homeorhesis, Acta Biotheoretica, 54(2) 61-101. [3] K. Psiuk-Maksymowicz and E. Mamontov, 2006, The homeorhesis-based modelling and fast numerical analysis for oncogenic hyperplasia under radiotherapy, Mathl Comput. Modelling, Special Issue
|
|
| 9. |
|
|
| 10. |
- Romu, Thobias, et al.
(författare)
-
MANA - Multi scale adaptive normalized averaging
- 2011
-
Ingår i: 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro. - : IEEE conference proceedings. - 9781424441280 ; , s. 361-364
-
Konferensbidrag (refereegranskat)abstract
- It is possible to correct intensity inhomogeneity in fat–water Magnetic Resonance Imaging (MRI) by estimating a bias field based on the observed intensities of voxels classified as the pure adipose tissue. The same procedure can also be used to quantify fat volume and its distribution which opens up for new medical applications. The bias field estimation method has to be robust since pure fat voxels are irregularly located and the density varies greatly within and between image volumes. This paper introduces Multi scale Adaptive Normalized Average (MANA) that solves this problem bybasing the estimate on a scale space of weighted averages. By usingthe local certainty of the data MANA preserves details where the local data certainty is high and provides realistic values in sparse areas.
|
|
