Wyniki wyszukiwania - BazTech

1

U-Net based frames partitioning and volumetric analysis for kidney detection in tomographic images

Les Tomasz

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2021

|

Vol. 69, nr 3

art. no. e137051

EN

This work presents an automatic system for generating kidney boundaries in computed tomography (CT) images. This paper presents the main points of medical image processing, which are the parts of the developed system. The U-Net network was used for image segmentation, which is now widely used as a standard solution for many medical image processing tasks. An innovative solution for framing the input data has been implemented to improve the quality of the learning data as well as to reduce the size of the data. Precision-recall analysis was performed to calculate the optimal image threshold value. To eliminate false-positive errors, which are a common issue in segmentation based on neural networks, the volumetric analysis of coherent areas was applied. The developed system facilitates a fully automatic generation of kidney boundaries as well as the generation of a three-dimensional kidney model. The system can be helpful for people who deal with the analysis of medical images, medical specialists in medical centers, especially for those who perform the descriptions of CT examination. The system works fully automatically and can help to increase the accuracy of the performed medical diagnosis and reduce the time of preparing medical descriptions.

2

Segmentation of bone structures with the use of deep learning techniques

Krawczyk Zuzanna, Starzyński Jacek

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2021

|

Vol. 69, nr 3

art. no. e136751

EN

The paper is focused on automatic segmentation task of bone structures out of CT data series of pelvic region. The authors trained and compared four different models of deep neural networks (FCN, PSPNet, U-net and Segnet) to perform the segmentation task of three following classes: background, patient outline and bones. The mean and class-wise Intersection over Union (IoU), Dice coefficient and pixel accuracy measures were evaluated for each network outcome. In the initial phase all of the networks were trained for 10 epochs. The most exact segmentation results were obtained with the use of U-net model, with mean IoU value equal to 93.2%. The results where further outperformed with the U-net model modification with ResNet50 model used as the encoder, trained by 30 epochs, which obtained following result: mIoU measure – 96.92%, “bone” class IoU – 92.87%, mDice coefficient – 98.41%, mDice coefficient for “bone” – 96.31%, mAccuracy – 99.85% and Accuracy for “bone” class – 99.92%.

3

Classification of high resolution satellite images using improved U-Net

Wang Yong, Zhang Dongfang, Dai Guangming

International Journal of Applied Mathematics and Computer Science

|

2020

|

Vol. 30, no. 3

399--413

EN

Satellite image classification is essential for many socio-economic and environmental applications of geographic information systems, including urban and regional planning, conservation and management of natural resources, etc. In this paper, we propose a deep learning architecture to perform the pixel-level understanding of high spatial resolution satellite images and apply it to image classification tasks. Specifically, we augment the spatial pyramid pooling module with image-level features encoding the global context, and integrate it into the U-Net structure. The proposed model solves the problem consisting in the fact that U-Net tends to lose object boundaries after multiple pooling operations. In our experiments, two public datasets are used to assess the performance of the proposed model. Comparison with the results from the published algorithms demonstrates the effectiveness of our approach.

4

BCT Boost Segmentation with U-net in TensorFlow

Wieczorek Grzegorz, Antoniuk Izabella, Kurek Jarosław, Świderski Bartosz, Kruk Michał, Pach Jakub, Orłowski Arkadiusz

Machine Graphics and Vision

|

2019

|

Vol. 28, No. 1/4

25--34

EN

In this paper we present a new segmentation method meant for boost area that remains after removing the tumour using BCT (breast conserving therapy). The selected area is a region on which radiation treatment will later be made. Consequently, an inaccurate designation of this region can result in a treatment missing its target or focusing on healthy breast tissue that otherwise could be spared. Needless to say that exact indication of boost area is an extremely important aspect of the entire medical procedure, where a better definition can lead to optimizing of the coverage of the target volume and, in result, can save normal breast tissue. Precise definition of this area has a potential to both improve the local control of the disease and to ensure better cosmetic outcome for the patient. In our approach we use U-net along with Keras and TensorFlow systems to tailor a precise solution for the indication of the boost area. During the training process we utilize a set of CT images, where each of them came with a contour assigned by an expert. We wanted to achieve a segmentation result as close to given contour as possible. With a rather small initial data set we used data augmentation techniques to increase the number of training examples, while the final outcomes were evaluated according to their similarity to the ones produced by experts, by calculating the mean square error and the structural similarity index (SSIM).