Automatic bladder segmentation from CT images using deep CNN and 3D fully connected CRF-RNN

Xu, Xuanang; Zhou, Fugen; Liu, Bo

doi:10.1007/s11548-018-1733-7

Cited by 37 publications

(35 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Ref. , the authors used a small test set of 24 cases while in our study the test set consisted of 92 cases. The bladders in Ref.…”

Section: Discussionmentioning

confidence: 99%

“…The bladders in Ref. were imaged with noncontrast‐enhanced CT, while in our experiments the bladders were partially or entirely filled with excreted contrast material or without any contrast material, which makes the segmentation more challenging as discussed above. We did not apply preprocessing of the CT images (such as enhancement density filters) or contour refinement of the U‐DL contour.…”

Section: Discussionmentioning

confidence: 99%

“…The DCNN‐LS was superior to many gradient‐based segmentation methods such as CLASS, but inaccurate segmentation occurred in some cases that had poor image quality or in cases where advanced bladder cancer had spread into neighboring organs. A recent study reported the training of a deep learning‐based model that involved a convolutional neural network (CNN) and a three‐dimensional (3D) fully connected conditional random field recurrent neural network (CRF‐RNN) to perform bladder segmentation using 100 training and 24 test CT images.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

U‐Net based deep learning bladder segmentation in CT urography

Hadjiiski

Wei

et al. 2019

Medical Physics

View full text Add to dashboard Cite

Objectives: To develop a U-Net-based deep learning approach (U-DL) for bladder segmentation in computed tomography urography (CTU) as a part of a computer-assisted bladder cancer detection and treatment response assessment pipeline. Materials and methods: A dataset of 173 cases including 81 cases in the training/validation set (42 masses, 21 with wall thickening, 18 normal bladders), and 92 cases in the test set (43 masses, 36 with wall thickening, 13 normal bladders) were used with Institutional Review Board approval. An experienced radiologist provided three-dimensional (3D) hand outlines for all cases as the reference standard. We previously developed a bladder segmentation method that used a deep learning convolution neural network and level sets (DCNN-LS) within a user-input bounding box. However, some cases with poor image quality or with advanced bladder cancer spreading into the neighboring organs caused inaccurate segmentation. We have newly developed an automated U-DL method to estimate a likelihood map of the bladder in CTU. The U-DL did not require a user-input box and the level sets for postprocessing. To identify the best model for this task, we compared the following models: (a) two-dimensional (2D) U-DL and 3D U-DL using 2D CT slices and 3D CT volumes, respectively, as input, (b) U-DLs using CT images of different resolutions as input, and (c) U-DLs with and without automated cropping of the bladder as an image preprocessing step. The segmentation accuracy relative to the reference standard was quantified by six measures: average volume intersection ratio (AVI), average percent volume error (AVE), average absolute volume error (AAVE), average minimum distance (AMD), average Hausdorff distance (AHD), and the average Jaccard index (AJI). As a baseline, the results from our previous DCNN-LS method were used. Results: In the test set, the best 2D U-DL model achieved AVI, AVE, AAVE, AMD, AHD, and AJI values of 93.4 AE 9.5%, À4.2 AE 14.2%, 9.2 AE 11.5%, 2.7 AE 2.5 mm, 9.7 AE 7.6 mm, 85.0 AE 11.3%, respectively, while the corresponding measures by the best 3D U-DL were 90.6 AE 11.9%, À2.3 AE 21.7%, 11.5 AE 18.5%, 3.1 AE 3.2 mm, 11.4 AE 10.0 mm, and 82.6 AE 14.2%, respectively. For comparison, the corresponding values obtained with the baseline method were 81.9 AE 12.1%, 10.2 AE 16.2%, 14.0 AE 13.0%, 3.6 AE 2.0 mm, 12.8 AE 6.1 mm, and 76.2 AE 11.8%, respectively, for the same test set. The improvement for all measures between the best U-DL and the DCNN-LS were statistically significant (P < 0.001). Conclusion: Compared to a previous DCNN-LS method, which depended on a user-input bounding box, the U-DL provided more accurate bladder segmentation and was more automated than the previous approach.

show abstract

“…In Ref. , the authors used a small test set of 24 cases while in our study the test set consisted of 92 cases. The bladders in Ref.…”

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

U‐Net based deep learning bladder segmentation in CT urography

Hadjiiski

Wei

et al. 2019

Medical Physics

View full text Add to dashboard Cite

show abstract

“…20 Due to this characteristic, CNN has improved the accuracy of predicting various data types. Most of the results of prediction based on the CNN achieve a relatively high accuracy of 85% or above 21,22 ; however, it should be noted that CNN needs a large dataset for training from scratch. It is quite difficult to obtain huge datasets for medical images due to patient privacy issues.…”

Section: Discussionmentioning

confidence: 99%

Comparison of CT and MRI images for the prediction of soft-tissue sarcoma grading and lung metastasis via a convolutional neural networks model

Zhang

Ren

2020

Clinical Radiology

View full text Add to dashboard Cite

To realise the automated prediction of soft-tissue sarcoma (STS) grading and lung metastasis based on computed tomography (CT), T1-weighted (T1W) magnetic resonance imaging (MRI), and fat-suppressed T2-weighted MRI (FST2W) via the convolutional neural networks (CNN) model. MATERIALS AND METHODS: MRI and CT images of 51 patients diagnosed with STS were analysed retrospectively. The patients could be divided into three groups based on disease grading: high-grade group (n¼28), intermediate-grade group (n¼15), low-grade group (n¼8). Among these patients, 32 had lung metastasis, while the remaining 19 had no lung metastasis. The data were divided into the training, validation, and testing groups according to the ratio of 5:2:3. The receiver operating characteristic (ROC) curves and accuracy values were acquired using the testing dataset to evaluate the performance of the CNN model. RESULTS: For STS grading, the accuracy of the T1W, FST2W, CT, and the fusion of T1W and FST2W testing data were 0.86, 0.89, 0.86, and 0.85, respectively. In addition, Area Under Curve (AUC) were 0.96, 0.97, 0.97, and 0.94 respectively. For the prediction of lung metastasis, the accuracy of the T1W, FST2W, CT, and the fusion of T1W and FST2W test data were 0.92, 0.93, 0.88, and 0.91, respectively. The corresponding AUC values were 0.97, 0.96, 0.95, and 0.95, respectively. FST2W MRI performed best for predicting STS grading and lung metastasis. CONCLUSION: MRI and CT images combined with the CNN model can be useful for making predictions regarding STS grading and lung metastasis, thus providing help for patient diagnosis and treatment.

show abstract

“…Zheng and Yi [45] tweaked the traditional "vanilla RNN" architecture by adding a "competitive layer model" to produce accurate results in brain MRI segmentation. Xu et al [46] applied a novel dual-channel preprocessing (composed of original CT image as well as an enhanced density map) method as well as a CRF-RNN (CRF stands for Conditional Random Fields) model combined with CNN to extract maximal performance for bladder segmentation from CT images. The hybrid deep learning model was better than using only CNNs, because some limitations of the CNN, as mentioned above, were rectified by combining the CRF-RNN model.…”

Section: B Recurrent Neural Network (Rnns)mentioning

confidence: 99%

The Rise of Deep Learning in Radiology: An Overview of Recent Research

Chatterjee¹

2019

IJRASET

View full text Add to dashboard Cite

This is a consolidated look at the applications of various deep learning techniques in the field of radiology. For the past few years, deep learning has pervaded every field and the deep learning revolution has opened up new frontiers in artificial intelligence. From healthcare to art or from education to business-it has been applied successfully in a range of different domains to attain state-of-the-art, often reporting near human-level performance. Hence, in the field of radiology too, especially for image interpretation tasks, deep learning techniques are being increasingly used in recent times to optimize the medical workflow and to achieve better patient care and efficient medical surveillance. Convolutional neural networks (CNNs) are mostly dominant in the case of image interpretation applications in radiology, because of their unprecedented success in image-related applications in other domains such as computer vision. However, other deep learning techniques like recurrent neural networks (RNNs) and generative adversarial networks (GANs) are also being used in recent research for various image-related tasks in radiology like classification, segmentation and detection. In this paper, the imaging modalities associated with this field and the application of different deep learning techniques to these have been discussed at length. Moreover, deep learning can also be applied to radiology use cases other than image interpretation, such as patient scheduling or the processing of free-text radiology reports to improve healthcare surveillance. Finally, in this study, the practical challenges as well as the future research directions of this domain have been discussed. Some challenges include dearth of annotated data, the fear of AI unseating radiologist professionals, legal and ethical issues, black box behaviour of neural networks and adversarial fooling of deep learning algorithms by reverse engineering. To counter some of these problems, a few trends in applying deep learning to radiology in the future may include improved visualisation techniques, integration of the entire workflow for practical usability, unsupervised methods of deep learning like auto-encoders and improving research on GANs in radiology. This study may prove useful for researchers applying deep learning to various radiology use cases by providing a detailed overview of the state-of-theart research in the field.

show abstract

Automatic bladder segmentation from CT images using deep CNN and 3D fully connected CRF-RNN

Cited by 37 publications

References 21 publications

U‐Net based deep learning bladder segmentation in CT urography

U‐Net based deep learning bladder segmentation in CT urography

Comparison of CT and MRI images for the prediction of soft-tissue sarcoma grading and lung metastasis via a convolutional neural networks model

The Rise of Deep Learning in Radiology: An Overview of Recent Research

Contact Info

Product

Resources

About