Accurate and robust deep learning-based segmentation of the prostate clinical target volume in ultrasound images

Karimi, Davood; Zeng, Qi; Mathur, Prateek; Avinash, Apeksha; Mahdavi, Sara; Spadinger, Ingrid; Abolmaesumi, Purang; Salcudean, Septimiu E.

doi:10.1016/j.media.2019.07.005

Cited by 96 publications

(48 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The present study uses a network architecture called 3D U-net, which has been used in previous studies involving deep learning in brachytherapy. 6,18,27,30,36 The U-net structure has been used for segmentation of both needles and organs such as the prostate. 18,30 A similar structure called V-net has also been proposed, and this too has been used for segmentation of the prostate.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning‐based digitization of prostate brachytherapy needles in ultrasound images

et al. 2020

View full text Add to dashboard Cite

To develop, and evaluate the performance of, a deep learning-based three-dimensional (3D) convolutional neural network (CNN) artificial intelligence (AI) algorithm aimed at finding needles in ultrasound images used in prostate brachytherapy. Methods: Transrectal ultrasound (TRUS) image volumes from 1102 treatments were used to create a clinical ground truth (CGT) including 24422 individual needles that had been manually digitized by medical physicists during brachytherapy procedures. A 3D CNN U-net with 128 × 128 × 128 TRUS image volumes as input was trained using 17215 needle examples. Predictions of voxels constituting a needle were combined to yield a 3D linear function describing the localization of each needle in a TRUS volume. Manual and AI digitizations were compared in terms of the root-mean-square distance (RMSD) along each needle, expressed as median and interquartile range (IQR). The method was evaluated on a data set including 7207 needle examples. A subgroup of the evaluation data set (n = 188) was created, where the needles were digitized once more by a medical physicist (G1) trained in brachytherapy. The digitization procedure was timed. Results: The RMSD between the AI and CGT was 0.55 (IQR: 0.35-0.86) mm. In the smaller subset, the RMSD between AI and CGT was similar (0.52 [IQR: 0.33-0.79] mm) but significantly smaller (P < 0.001) than the difference of 0.75 (IQR: 0.49-1.20) mm between AI and G1. The difference between CGT and G1 was 0.80 (IQR: 0.48-1.18) mm, implying that the AI performed as well as the CGT in relation to G1. The mean time needed for human digitization was 10 min 11 sec, while the time needed for the AI was negligible. Conclusions: A 3D CNN can be trained to identify needles in TRUS images. The performance of the network was similar to that of a medical physicist trained in brachytherapy. Incorporating a CNN for needle identification can shorten brachytherapy treatment procedures substantially.

show abstract

Section: Discussionmentioning

confidence: 99%

“…Previous studies include work on digitizing gynaecological applicators, 6,26,27 prostate seeds, 28,29 and segmentation of the prostate boundaries. [30][31][32] Recent publications by Zhang et al 18 , Wang et al 33 and Dise et al 26 have used deep learning for identifying needles in 3D TRUS images.…”

Section: Discussionmentioning

confidence: 99%

Deep learning‐based digitization of prostate brachytherapy needles in ultrasound images

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Manual segmentation of the prostate on TRUS imaging is time-consuming and often not reproducible. For these reasons, several studies have applied deep learning to automatically segment the prostate using TRUS imaging [25][26][27][28][29][30][31].…”

Section: Challenges Applying Deep Learning To Abdominal Us Imagingmentioning

confidence: 99%

Current status of deep learning applications in abdominal ultrasonography

Song

2021

Ultrasonography

View full text Add to dashboard Cite

Ultrasonography. 2020 Sep 2. Epub ahead of print Deep learning is one of the most popular artificial intelligence techniques used in the medical field. Although it is at an early stage compared to deep learning analyses of computed tomography or magnetic resonance imaging, studies applying deep learning to ultrasound imaging have been actively conducted. This review analyzes recent studies that applied deep learning to ultrasound imaging of various abdominal organs and explains the challenges encountered in these applications.

show abstract

“…To that goal, we replaced the output layer with a new layer consisting of only a single neuron. (2) We reduced the output size by a factor of 4 by replacing the output layer with a new layer consisting of 16 neurons to see if higher performance is achievable by reducing the number of parameters. In this case, aberrator profiles were downsampled to a vector of size 16 during training the network.…”

Section: ) Output Sizementioning

confidence: 99%

Phase Aberration Correction: A Convolutional Neural Network Approach

2020

View full text Add to dashboard Cite

One of the main sources of image degradation in ultrasound imaging is the phase aberration effect, which imposes limitations to both data acquisition and reconstruction. Phase aberration is induced by spatial changes in sound velocity compared to the default values and degrades the quality of beam focusing. In addition, it prevents received channel signals to be summed coherently. In this paper, for the first time, we propose a method to estimate the aberrator profile from an ultrasound B-mode image using a deep convolutional neural network (CNN) in order to compensate for the phase aberration effect. In contrast to traditional methods, which mostly apply time-consuming processing techniques on channel RF signals and need several iterations for a reasonable accuracy, the proposed approach is computationally efficient, and utilizes only the B-mode image to estimate the aberrator profile in one shot with a high accuracy. We experimentally investigate main characteristics of the proposed approach and present a quantitative evaluation of the estimated aberrator profile. The proposed method is compared with the conventional delay-and-sum (DAS) method and a method based on nearest-neighbor cross-correlation (NNCC). Results demonstrate that the proposed CNN method substantially outperforms other methods. INDEX TERMS Convolutional neural networks, deep learning, phase aberration, ultrasound.

show abstract

Accurate and robust deep learning-based segmentation of the prostate clinical target volume in ultrasound images

Cited by 96 publications

References 22 publications

Deep learning‐based digitization of prostate brachytherapy needles in ultrasound images

Deep learning‐based digitization of prostate brachytherapy needles in ultrasound images

Current status of deep learning applications in abdominal ultrasonography

Phase Aberration Correction: A Convolutional Neural Network Approach

Contact Info

Product

Resources

About