Automatic intraprostatic lesion segmentation in multiparametric magnetic resonance images with proposed multiple branch UNet

Chen, Yizheng; Li, Xing; Yu, Lequan; Bagshaw, H.P.; Buyyounouski, Mark K.; Han, Bin

doi:10.1002/mp.14517

Cited by 29 publications

(31 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This study provided an automatic approach for this image registration task, and the adopted deformable registration method could help to correct the difference of the organ shapes. In a previous study, 34 we worked on the automatic segmentation of intraprostatic lesions on MR images, and then the lesion contours could be automatically propagated to the planning US images during the HDR brachytherapy procedure, allowing for dose escalation to specific intraprostatic lesions if warranted.…”

Section: Discussionmentioning

confidence: 99%

“…As for the MR dataset for segmentation, we utilized our previous database originally for intraprostatic lesion segmentation, 34 in which the prostate and intraprostatic lesions were manually contoured on multi‐modality MR images. The T2W images with prostate contours of 121 patient cases in this database were used for the prostate segmentation on MR images in this study.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

MR to ultrasound image registration with segmentation‐based learning for HDR prostate brachytherapy

Chen

et al. 2021

Medical Physics

Self Cite

View full text Add to dashboard Cite

Purpose Propagation of contours from high‐quality magnetic resonance (MR) images to treatment planning ultrasound (US) images with severe needle artifacts is a challenging task, which can greatly aid the organ contouring in high dose rate (HDR) prostate brachytherapy. In this study, a deep learning approach was developed to automatize this registration procedure for HDR brachytherapy practice. Methods Because of the lack of training labels and difficulty of accurate registration from inferior image quality, a new segmentation‐based registration framework was proposed for this multi‐modality image registration problem. The framework consisted of two segmentation networks and a deformable registration network, based on the weakly ‐supervised registration strategy. Specifically, two 3D V‐Nets were trained for the prostate segmentation on the MR and US images separately, to generate the weak supervision labels for the registration network training. Besides the image pair, the corresponding prostate probability maps from the segmentation were further fed to the registration network to predict the deformation matrix, and an augmentation method was designed to randomly scale the input and label probability maps during the registration network training. The overlap between the deformed and fixed prostate contours was analyzed to evaluate the registration accuracy. Three datasets were collected from our institution for the MR and US image segmentation networks, and the registration network learning, which contained 121, 104, and 63 patient cases, respectively. Results The mean Dice similarity coefficient (DSC) results of the two prostate segmentation networks are 0.86 ± 0.05 and 0.90 ± 0.03, for MR images and the US images after the needle insertion, respectively. The mean DSC, center‐of‐mass (COM) distance, Hausdorff distance (HD), and averaged symmetric surface distance (ASSD) results for the registration of manual prostate contours were 0.87 ± 0.05, 1.70 ± 0.89 mm, 7.21 ± 2.07 mm, 1.61 ± 0.64 mm, respectively. By providing the prostate probability map from the segmentation to the registration network, as well as applying the random map augmentation method, the evaluation results of the four metrics were all improved, such as an increase in DSC from 0.83 ± 0.08 to 0.86 ± 0.06 and from 0.86 ± 0.06 to 0.87 ± 0.05, respectively. Conclusions A novel segmentation‐based registration framework was proposed to automatically register prostate MR images to the treatment planning US images with metal artifacts, which not only largely saved the labor work on the data preparation, but also improved the registration accuracy. The evaluation results showed the potential of this approach in HDR prostate brachytherapy practice.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

MR to ultrasound image registration with segmentation‐based learning for HDR prostate brachytherapy

Chen

et al. 2021

Medical Physics

Self Cite

View full text Add to dashboard Cite

show abstract

“…The U-Net algorithm has been widely used for organ and lesion segmentation on MRI images, such as prostate [22,23] and prostate lesions [24]. However, this method is not easily applied to LN lesions.…”

Section: Discussionmentioning

confidence: 99%

Development and External Validation of the 3D U-Net Algorithm for Segmentation of Pelvic Lymph Nodes On Diffusion-Weighted Images

Liu

Sun

Han

et al. 2021

Preprint

View full text Add to dashboard Cite

Background: The 3D U-Net model has been proved to perform well in the automatic organ segmentation. The aim of this study is to evaluate the feasibility of the 3D U-Net algorithm for the automated detection and segmentation of lymph nodes (LNs) on pelvic diffusion-weighted imaging (DWI) images.Methods: A total of 393 DWI images of patients suspected of having prostate cancer (PCa) between January 2019 and December 2020 were collected for model development. Seventy-seven DWI images from another group of PCa patients imaged between January 2021 and April 2021 were collected for external validation. Segmentation performance was assessed using the Dice score, positive predictive value (PPV), true positive rate (TPR), and volumetric similarity (VS) with manual annotation of pelvic LNs as the reference. The accuracy with which the suspicious metastatic LNs (short diameter > 0.8 cm) were detected was evaluated using the area under the curve (AUC) at the patient level, and the precision, recall, and F1-score were determined at the lesion level. The consistency of LN staging on an external dataset between the model and radiologist was assessed using Cohen’s kappa coefficient.Results: In the testing set used for model development, the Dice score, TPR, PPV, and VS for the segmentation of suspicious LNs were 0.85, 0.82, 0.80 and 0.86, respectively. The precision, recall, and F1-score for the detection of suspicious LNs were 0.97, 0.98 and 0.97, respectively. In the external validation dataset, the AUC of the model for identifying PCa patients with suspicious LNs was 0.963 (95% CI: 0.892-0.993). High consistency of LN staging (Kappa = 0.922) was achieved between the model and expert radiologist.Conclusion: The 3D U-Net algorithm can accurately detect and segment pelvic LNs based on DWI images.

show abstract

“…Chen et. al [17] aimed to perform the segmentation of prostate lesions by using MR images. For this purpose, researchers proposed a novel segmentation network based on U-Net architecture design.…”

Section: Related Workmentioning

confidence: 99%

Effect of Preprocessing and Loss Functions on a Novel Dataset: An Application on MRI-Based Prostate Lesion Segmentation using U-Net

GURKAN

Budak²,

Karataş³

et al. 2022

Preprint

View full text Add to dashboard Cite

Prostate cancer is one of the riskiest types of cancer, while it is the second most common type of cancer in men and it is likely to result in death. While the diagnosis of prostate cancer is vital, it is complex, time-consuming, and the technologies used in its diagnosis are limited. The need for high technology for the diagnosis of prostate cancer is increasing day by day. One of the technologies used in health nowadays is artificial intelligence. Therefore, artificial intelligence-supported decision support systems should be used in the diagnosis of prostate cancer. However, the number of studies using artificial intelligence for the diagnosis of prostate cancer is limited. One of the most important reasons for this is the fewness of open-source data sets. Therefore, in this study, a novel MRI data set was created by an expert radiologist. The U-Net model was trained and tested using this data set. Different preprocessing methods and loss functions were studied for the training of the model. The used preprocessing techniques are (i) min-max normalization technique, (ii) z-score normalization technique, (iii) min-max normalization technique on 256 by 256 pixels cropped images, (iv) z-score normalization technique on 256 by 256 pixels cropped images, and (v) patch-wise analysis on 256 by 256 pixels cropped images. The used loss functions are Cross-Entropy, Dice, IoU, Tversky, Focal Tversky loss functions, and compound loss functions of these. Considering the combination of preprocessing techniques and loss functions, the compound loss function of Cross-Entropy and Dice loss functions have outperformed the patch-wise analysis on 256 by 256 pixels cropped images. In this pipeline, the U-Net model achieved a 0.917 dice score, 0.846 IoU score, 0.890 sensitivity, 0.998 specificity, 0.945 precision, and 0.995 accuracy. Consequently, a decision support system was developed for the diagnosis of prostate cancer.

show abstract

Automatic intraprostatic lesion segmentation in multiparametric magnetic resonance images with proposed multiple branch UNet

Cited by 29 publications

References 36 publications

MR to ultrasound image registration with segmentation‐based learning for HDR prostate brachytherapy

MR to ultrasound image registration with segmentation‐based learning for HDR prostate brachytherapy

Development and External Validation of the 3D U-Net Algorithm for Segmentation of Pelvic Lymph Nodes On Diffusion-Weighted Images

Effect of Preprocessing and Loss Functions on a Novel Dataset: An Application on MRI-Based Prostate Lesion Segmentation using U-Net

Contact Info

Product

Resources

About