Phase2Phase

Eldeniz, Cihat; Gan, Weijie; Chen, Sihao; Fraum, Tyler J.; Ludwig, Daniel; Yan, Yan; Liu, Jiaming; Vahle, Thomas; Krishnamurthy, Uday; Kamilov, Ulugbek S.; An, Hongyu

doi:10.1097/rli.0000000000000792

Cited by 17 publications

(31 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As well as image noise, motion artifacts are problematic in other MR sequences of the liver such as dynamic contrast-enhanced MRI. 20 Several techniques to improve image quality have been developed, [21][22][23] and reducing acquisition time with DLR might be also useful in other MR sequences of the liver.…”

Section: Discussionmentioning

confidence: 99%

Usefulness of Breath-Hold Fat-Suppressed T2-Weighted Images With Deep Learning–Based Reconstruction of the Liver

et al. 2022

View full text Add to dashboard Cite

Objectives: The aim of this study was to evaluate the usefulness of breath-hold turbo spin echo with deep learning-based reconstruction (BH-DL-TSE) in acquiring fat-suppressed T2-weighted images (FS-T2WI) of the liver by comparing this method with conventional free-breathing turbo spin echo (FB-TSE) and breath-hold half Fourier single-shot turbo spin echo with deep learning-based reconstruction (BH-DL-HASTE). Materials and Methods: The study cohort comprised 111 patients with suspected liver disease who underwent 3 T magnetic resonance imaging. Fifty-eight focal solid liver lesions ≥10 mm were also evaluated. Three sets of FS-T2WI were acquired using FB-TSE, prototypical BH-DL-TSE, and prototypical BH-DL-HASTE, respectively. In the qualitative analysis, 2 radiologists evaluated the image quality using a 5-point scale. In the quantitative analysis, we calculated the lesionto-liver signal intensity ratio (LEL-SIR). Friedman test and Dunn multiple comparison test were performed to assess differences among 3 types of FS-T2WI with respect to image quality and LEL-SIR. Results: The mean acquisition time was 4 minutes and 43 seconds ± 1 minute and 21 seconds (95% confidence interval, 4 minutes and 28 seconds to 4 minutes and 58 seconds) for FB-TSE, 40 seconds for BH-DL-TSE, and 20 seconds for BH-DL-HASTE. In the qualitative analysis, BH-DL-HASTE resulted in the fewest respiratory motion artifacts ( P < 0.0001). BH-DL-TSE and FB-TSE exhibited significantly less motion-related signal loss and clearer intrahepatic vessels than BH-DL-HASTE ( P < 0.0001). Regarding the edge sharpness of the left lobe, BH-DL-HASTE scored the highest ( P < 0.0001), and BH-DL-TSE scored higher than FB-TSE ( P = 0.0290). There were no significant differences among 3 types of FS-T2WI with respect to the edge sharpness of the right lobe ( P = 0.1290), lesion conspicuity ( P = 0.5292), and LEL-SIR ( P = 0.6026). Conclusions: BH-DL-TSE provides a shorter acquisition time and comparable or better image quality than FB-TSE, and could replace FB-TSE in acquiring FS-T2WI of the liver. BH-DL-TSE and BH-DL-HASTE have their own advantages and may be used complementarily.

show abstract

Section: Discussionmentioning

confidence: 99%

Usefulness of Breath-Hold Fat-Suppressed T2-Weighted Images With Deep Learning–Based Reconstruction of the Liver

et al. 2022

View full text Add to dashboard Cite

show abstract

“…In fact, up to the motion-compensated XD-MBDL step, iMoCo and our approach are essentially the same with both approaches relying on iterative reconstruction followed by motion field estimation. Several works 23,25 take a different approach and build motion compensated reconstruction workflows based entirely on deep learning architectures with the benefit of very fast compute times during inference relative to iterative methods. We chose to use iterative methods for motion field estimation over DL approaches because the performance of iterative approaches is well characterized and allows for instance-wise motion estimation that can be applied to a wide range of respiratory motion patterns.…”

Section: Xd-mbdl Image Quality Comparisonmentioning

confidence: 99%

“…10,20 Work incorporating non-rigid motion field estimation into these reconstructions has demonstrated even higher quality results as aligning data spatially across frames reduces temporal variations. 21,22 There are several recent works that explore incorporating spatiotemporal correlations across frames and motion correction into deep learning frameworks to improve reconstruction quality [23][24][25] These works, however, use Cartesian data that can be divided into smaller subsets for efficient training and thus do not have to manage the memory or training time constraints seen when reconstructing 3D non-Cartesian data. Furthermore, these works focus on either MBDL-based supervised training for image reconstruction approaches or purely image space noise2noise approaches as opposed to self-supervised training.…”

Section: Introductionmentioning

confidence: 99%

Motion compensated self supervised deep learning for highly accelerated 3D ultrashort Echo time pulmonary MRI

Miller

Johnson

2023

Magnetic Resonance in Med

View full text Add to dashboard Cite

To investigate motion compensated, self-supervised, model based deep learning (MBDL) as a method to reconstruct free breathing, 3D pulmonary UTE acquisitions. Theory and Methods:A self-supervised eXtra dimension MBDL architecture (XD-MBDL) was developed that combined respiratory states to reconstruct a single high-quality 3D image. Non-rigid motion fields were incorporated into this architecture by estimating motion fields from a lower resolution motion resolved (XD-GRASP) reconstruction. Motion compensated XD-MBDL was evaluated on lung UTE datasets with and without contrast and compared to constrained reconstructions and variants of self-supervised MBDL that do not account for dynamic respiratory states or leverage motion correction.Results: Images reconstructed using XD-MBDL demonstrate improved image quality as measured by apparent SNR (aSNR), contrast to noise ratio (CNR), and visual assessment relative to self-supervised MBDL approaches that do not account for dynamic respiratory states, XD-GRASP and a recently proposed motion compensated iterative reconstruction strategy (iMoCo). Additionally, XD-MBDL reduced reconstruction time relative to both XD-GRASP and iMoCo. Conclusion:A method was developed to allow self-supervised MBDL to combine multiple respiratory states to reconstruct a single image. This method was combined with graphics processing unit (GPU)-based image registration to further improve reconstruction quality. This approach showed promising results reconstructing a user-selected respiratory phase from free breathing 3D pulmonary UTE acquisitions.

show abstract

“…Malhotra [2] summarized the most commonly used image segmentation methods in the medical eld from 2017 to 2021, including CNN [3], deep belief network [4], R-CNN [5], V-Net [6], U-Net [7] and DeepLab [8]. On this basis, scholars have widely applied it to heart [9,10], brain [11,12], lung [13,14], liver [15,16], reproductive system [17], digestive system [18], chest [19], kidney [20] and eye [21][22][23] and other tissues and organs. They are committed to quantifying image data into digital indicators, and processing diversi ed imaging forms, so as to further explore hidden indicators bene cial to survival analysis.…”

Section: Introductionmentioning

confidence: 99%

Improving the prognostic prediction accuracy on breast cancer patients through comprehensive machine survival models

Kuang

Zhang

Chen

2023

Preprint

View full text Add to dashboard Cite

Background Breast cancer is known as one of the high-risk malignant tumors. Our previous studies have shown that computer-assisted quantification of large-scale tumor associated collagen signaling is an important prognostic indicator of breast cancer. However, the nonlinear relationship between the data has not been effectively mined, resulting in insufficient accuracy in prognosis prediction. Methods In this retrospective and multicenter study, we included 995 patients with invasive breast cancer and divided them into three cohorts, training cohort (N=438), internal validation cohort (N=293) and external validation cohort (N=264) respectively. Firstly, we used COX and random survival forest (SRF) to explore the significance of variables. And then we combined the survival models with machine learning into eight comprehensive machine survival models in order to improve the prognosis of breast cancer. Finally, 10 new survival indicators generated from the above models were used to classify patients into low risk and high risk under Kaplan-Meier method and Log-rank test. Result The baseline data and TACS of patients not only show their own non-linear, but also have a strong nonlinear complementary effect between them, and they even interact and promote each other. It is worth noting that the effect of the comprehensive machine survival models are better than that of the clinical model (CLI). Specifically, except SRF-2 does it best in the training cohort (AUC: 0.956 [0.904,0.973]), SRFNET has the best performance in the internal validation cohort (AUC: 0.961 [0.941,0.978]) and the external validation cohort (AUC: 0.935 [0.891,0.967]). Meanwhile, 10 new survival indicators generated from the above comprehensive machine survival models are not only significantly related to DFS, but also able to effectively distinguish high and low risk patients (P < 0.0001) based on the cutoff value of the 5-year time point. Conclusion The comprehensive machine survival models proposed in this paper improve the recurrence prediction accuracy of breast cancer patients by mining the nonlinearity between patients’ characteristic, so that we will provide effective tools and reliable basis for doctors to carry out precise treatment.

show abstract

Phase2Phase

Abstract: Supplemental digital content is available in the text.

Cited by 17 publications

References 40 publications

Usefulness of Breath-Hold Fat-Suppressed T2-Weighted Images With Deep Learning–Based Reconstruction of the Liver

Usefulness of Breath-Hold Fat-Suppressed T2-Weighted Images With Deep Learning–Based Reconstruction of the Liver

Motion compensated self supervised deep learning for highly accelerated 3D ultrashort Echo time pulmonary MRI

Improving the prognostic prediction accuracy on breast cancer patients through comprehensive machine survival models

Contact Info

Product

Resources

About