xiang liu¹ and xiaoying wang^1
1peking university first hospital, Beijing, China

Keywords: Liver, Cancer

This retrospective study aims to develop an automated algorithm for segmentation of liver metastases based on a deep learning method and assess its efficacy for treatment response assessment according to the RECIST 1.1 criteria. One hundred and sixteen treated patients with clinically confirmed liver metastases were enrolled. A 3D U-Net algorithm was trained for automated liver metastases segmentation and treatment response assessment. The results demonstrated that the automated liver metastases segmentation was capable of evaluating treatment response, with comparable results to the junior radiologist and superior to that of the fellow radiologist.

View Presentation Video

Hreedi Dev¹, Chenglin Zhu¹, Arman Sharbatdaran¹, Syed I. Raza¹, Sophie Wang¹, Dominick J. Romano¹, Akshay Goel¹, Kurt Teichman¹, Mina C. Moghadam¹, George Shih¹, Jon D. Blumenfeld^1,2, Daniil Shimonov², James Chevalier², and Martin R. Prince^1,3
1Radiology, Weill Cornell Medicine, New York, NY, United States, ²Rogosin Institute, New York, NY, United States, ³Radiology, Columbia University Vagelos College of Physicians and Surgeons, New York, NY, United States

Keywords: Kidney, Kidney, MRI, ADPKD

Organ volume measurements on MRI are typically performed on a single pulse sequence because of the tedious process of manual contouring. Here we use deep learning to automate kidney segmentations so that kidney volume can be measured on five abdominal MRI sequences.  In 17 subjects scanned twice within 3 weeks (when no change in kidney volume was expected), the power of averaging 5 measurements improved reproducibility, achieving 2.5% absolute percent difference compared to 5.9% with manual contouring (p<0.05). Absolute percent error was reduced further to 2.1%, p<0.05 compared to manual segmentations, by excluding outlier measurements.

Qiurong Wei¹, Weicui Chen¹, Liting Mao¹, Kan Deng², Zhaoxian Yan¹, Weikang Huang¹, and Xian Liu^1
1Radiology, The Second Affiliated Hospital of Guangzhou University of Chinese Medicine, Guangzhou, China, ²Philips Healthcare, Guangzhou, China

Keywords: Cancer, Radiomics

The response to neoadjuvant chemoradiotherapy (nCRT) in patients with locally advanced rectal cancer (LARC) is especially important for prognostic and management decisions. In this study, we used cross-vendor data from two centers to validate the generalization ability of radiomics model based on multiparametric-MRI (MP-MRI) for predicting pCR and to compare the discriminatory performance of different classifiers. Our results demonstrated that radiomics can be used to predict pCR. The clinical-radiomics model had superior performance compared to the radiomics model and clinical model. Furthermore, the RF classifier outperformed the other classifiers in prediction.

Pengyang Feng¹, Nan Meng², Xuan Yu², Yaping Wu², Fangfang Fu², Yu Luo², Han Jiang³, Ziqiang Li³, Jianmin Yuan⁴, Yang Yang⁵, Zhe Wang⁴, and Meiyun Wang*^2
1Department of Medical Imaging, Henan University People’s Hospital & Henan Provincial People’s Hospital, Zhengzhou, China, ²Department of Medical Imaging, Zhengzhou University People’s Hospital & Henan Provincial People’s Hospital, Zhengzhou, China, ³Department of Medical Imaging, Xinxiang Medical University Henan Provincial People’s Hospital, Zhengzhou, China, ⁴Central Research Institute, UIH Group, Shanghai, China, ⁵Beijing United Imaging Research Institute of Intelligent Imaging, UIH Group, Beijing, China

Keywords: Cancer, Radiomics

Three-dimensional ultrashort echo time (3D-UTE) is a novel MRI technique, which yields similar diagnostic results as conventional pulmonary computed tomography (CT). Our results showed that the predication model based on clinical factors and 3D-UTE radiomics features could noninvasively assess the subtype of in non-small cell lung cancer. Compared with the CT model, it has similar diagnostic efficiency but less radiation, which is expected to provide new ideas for related research.

View Presentation Video

Leon Bischoff^1,2, Alexander Isaak^1,2, Christoph Katemann³, Dmitrij Kravchenko^1,2, Narine Mesropyan^1,2, Christoph Endler^1,2, Barbara Wichtmann^1,2, Oliver Weber³, Johannes Peeters⁴, Claus Christian Pieper¹, Daniel Kütting^1,2, Alois Martin Sprinkart^1,2, Ulrike Attenberger¹, and Julian Luetkens^1,2
1Department of Diagnostic and Interventional Radiology, University Hospital Bonn, Bonn, Germany, ²Quantitative Imaging Lab, University Hospital Bonn, Bonn, Germany, ³Philips GmbH Market DACH, Hamburg, Germany, ⁴Philips MR Clinical Science, Best, Netherlands

Keywords: Prostate, Machine Learning/Artificial Intelligence, Prostate cancer

In this prospective study, 56 male patients with suspected prostate cancer were included to evaluate an artificial intelligence (AI) based reconstruction method for T2-weighted sequences in multiparametric MRI (mpMRI). After comparison with conventionally acquired and reconstructed sequences we found the new AI-based method to produce images with higher image sharpness and delineation of lesions, confirmed by both qualitative and quantitative analysis. This is accompanied by a reduction of scan time by 29-37%. Confidence in the assessed PI-RADS scores was significantly higher for the AI-reconstruction. This new technique could therefore potentially increase diagnostic accuracy of mpMRI of the prostate.

View Presentation Video

Ning Chien¹, Cheng-Ya Yeh¹, Yi-Chen Chen¹, Yeun-Chung Chang², Chia-Wei Li³, Chien-Yuan Lin³, Patricia Lan⁴, Xinzeng Wang⁵, Arnaud Guidon⁶, and Kao-Lang Liu^1
1Department of Medical Imaging, National Taiwan University Cancer Center and National Taiwan University College of Medicine, Taipei, Taiwan, ²Department of Medical Imaging, National Taiwan University Hospital and National Taiwan University College of Medicine, Taipei, Taiwan, ³GE Healthcare, Taipei, Taiwan, ⁴GE Healthcare, Menlo Park, CA, United States, ⁵GE Healthcare, Houston, TX, United States, ⁶GE Healthcare, Boston, MA, United States

Keywords: Breast, Machine Learning/Artificial Intelligence, Deep learning reconstruction, Multi-shot DWI

Diffusion-weighted imaging (DWI) in the breast is limited by image distortion, which can be improved with multi-shot DWI (MUSE). We conducted a pilot study to investigate the impact of deep-learning reconstruction (DLRecon) on MUSE image quality. Compared with the non-DL MUSE images, the MUSE DLRecon showed higher SNR without altering the mean ADC value. Moreover, the higher shots MUSE DL with reduced NEX could provide less-distortion DWI with comparable SNR and scan time to 2-shot MUSE imaging, which is commonly used in the clinical setting. Preliminary results indicate the feasibility of MUSE-DWI in the breast with higher number of shots.

View Presentation Video

Fangrong Liang^1,2, Wanli Zhang^1,2, Jiamin Li^1,2, Yongzhou Xu³, Aaron Zhang³, Xinqing Jiang^1,2, Xin Zhen⁴, and Ruimeng Yang^1,2
1Department of Radiology, The Second Affifiliated Hospital, School of Medicine, South China University of Technology, Guangzhou, China, ²Department of Radiology, Guangzhou First People’s Hospital, Guangzhou, China, ³Philips Healthcare, Guangzhou, China, ⁴School of Biomedical Engineering, Southern Medical University, Guangzhou, China

Keywords: Breast, Diffusion/other diffusion imaging techniques, Tumor proliferative burden; Triple negative breast cancer; Multiparametric magnetic resonance imaging

Triple negative breast cancer (TNBC) is highly heterogeneous, with poorer prognosis, higher recurrence rates and severe treatment challenges. Accurate preoperative identification of TNBC is helpful for individualized patient management. Based on multiparametric magnetic resonance imaging (mMRI), we employed whole-tumor ADC maps-based radiomics (R_ADC) model, tumor proliferative burden (TPB_ADC) model, mMRI-based feature fusion radiomics (R_FF) model and combinational R_FF-TPB_ADC model to investigate their performance in distinguishing TNBC from non-TNBC. Our results showed that the R_FF-TPB_ADC model outperformed the R_ADC, TPB_ADC, and R_FF models by integrating mMRI radiomics features and TPBs, demonstrating its potential for classification of breast cancer.

View Presentation Video

Zifeng Lian¹, Qiqi Lu¹, and Yanqiu Feng^1
1School of Biomedical Engineering, Southern Medical University, Guangzhou, China

Keywords: Relaxometry, Liver, T2*, liver iron content

MRI T2* relaxometry is a reliable method for assessing liver iron overload. To develop a fully automatic T2* relaxometry data processing method for assessing liver iron overload, we promoted a semi-automatic parenchyma extraction to an automatic approach by introducing a modified TransUNet on R2* map for the segmentation of whole liver. The proposed method showed excellent liver segmentation performance on the internal and external test sets and yielded T2* measurements highly consistent with those by the semi-automatic method. This fully automatic approach will enable an efficient and reliable measurement of liver T2* for assessing hepatic iron content in clinical practice.

View Presentation Video

Baoer Liu¹, Pingjing Wang², Jianbin Huang¹, Wu Zhou², and Yikai Xu^1
1Nanfang Hospital, Southern Medical University, Guangzhou, China, ²Guangzhou University of Chinese Medicine, Guangzhou, China

Keywords: Liver, Cancer, Gd-EOB-DTPA-enhanced MRI

An accurate preoperative assessment of microvascular invasion (MVI) in patients with hepatocellular carcinoma (HCC) is of great clinical importance in choosing appropriate surgical interventions. We aimed to investigate diagnostic performance of Gd-EOB-DTPA-enhanced MRI for prediction of MVI in HCC using convolutional neural network (CNN). The CNN model based on hepatobiliary phase (HBP) images had great diagnostic efficiency for the prediction of MVI with the AUC of 0.858 (range, 0.854, 0.893). Deep learning with CNN based on Gd-EOB-DTPA-enhanced MRI can be conducive to preoperative prediction of MVI in HCC.

Xianling Qian¹, Yunfei Zhang², Gengyun Miao¹, Yongming Dai², and Mengsu Zeng^1
1Zhongshan Hospital, Fudan University, Shanghai, China, ²MR Collaboration, Central Research Institute, United Imaging Healthcare, Shanghai, China

Keywords: Liver, Radiomics

Microvascular invasion (MVI) is a significant adverse prognostic indicator of intrahepatic cholangiocarcinoma (ICC), and affects the selection of individualized treatment regimen. However, the preoperative imaging-based identification of MVI status is rather difficult. This study sought to establish a multi-sequence and multi-scale MR image-based radiomics nomogram for the presurgical prediction of MVI in ICC.

View Presentation Video

Christopher Keen¹, Peter Šereš¹, Justin Grenier¹, Robert Stobbe¹, Ian Paterson², Kumar Punithakumar ³, Jacob Jaremko³, and Richard Thompson^1
1Department of Biomedical Engineering, University of Alberta, Edmonton, AB, Canada, ²Division of Cardiology, University of Alberta, Edmonton, AB, Canada, ³Department of Radiology and Diagnostic Imaging, University of Alberta, Edmonton, AB, Canada

Keywords: Machine Learning/Artificial Intelligence, Lung

Accurate segmentation is required to perform quantitative analysis on lung parenchyma in ultrashort echo time (UTE) proton MRI. Deep learning methods offer a solution to this problem, however, previous application to UTE lung MRI is limited. A deep learning model was trained to segment lung parenchyma using fine tuned region growing masks as a reference. To test the generalizability of the model the performance of three different 3D UTE k-space trajectories were compared. Overall, the model produced high quality segmentation for the different acquisition approaches with improvements over training data in areas of high vessel density and high signal intensity.

Lan He^1,2 and Zaiyi Liu^1,2
1Department of Radiology, Guangdong Provincial People's Hospital , Guangdong Academy of Medical Sciences, Guangzhou, China, ²Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China

Keywords: Lung, Radiomics, Prognosis prediction; Non-small cell lung cancer

Rad-molGPA index is prognostic for non-small cell lung cancer patients with brain metastases. The updated Rad-molGPA incorporating radiomics features into the Lung-molGPA is a user-friendly tool to facilitate clinical decision making and appropriate stratification of future clinical trials.

View Presentation Video

Xu Hua Gong¹, Lei Lv², and Li Jun Qian^1
1Radiology, Renji Hospital, Shanghai Jiao Tong University, School of Medicine, Shanghai, China, ²ShuKun (BeiJing) Technology Co., Ltd., Beijing, China

Keywords: Liver, Cancer

Quantitative European Association for the Study of the Liver (qEASL) can better evaluate the therapeutic efficacy of hepatocellular carcinoma (HCC) after transcatheter arterial chemoembolization (TACE). Conventional qEASL is often achieved semi-automatically, which is tedious, labor-intensive, and time-consuming. In this study, we provide and assess an automatic qEASL approach based on a VoxelMorph and Faster R-CNN framework named Shukun PortalDoc. We compared the consistency of the proposed method with the widely accepted semi-automatic software (MultiModality Tumor Tracking, Philips IntelliSpace Portal, Philips healthcare) in assessing qEASL of hepatocellular carcinoma after TACE.

View Presentation Video

Tom Syer¹, Nikolaos Dikaios², Thomas Parry³, Giorgio Brembilla³, Mrishta Brizmohun³, Saurabh Singh⁴, Susan Heavey⁵, Hayley Pye⁶, Hayley Whitaker⁵, Sue Mallett³, David Atkinson³, and Shonit Punwani^3
1Centre for Medical Imaging, University College London, London, United Kingdom, ²Mathematics Research Center, Academy of Athens, Athen, Greece, ³Centre for Medical Imaging, Univeristy College London, London, United Kingdom, ⁴Radiology, Univeristy College London Hospital, London, United Kingdom, ⁵Department of Targeted Intervention, Univeristy College London, London, United Kingdom, ⁶National Pathology Imaging Co-operative, Leeds, United Kingdom

Keywords: Prostate, Machine Learning/Artificial Intelligence

Machine learning for classifying prostate mpMRI lesions may help reduce unnecessary biopsies. However, external validation with multiple scanners and readers is required before the clinical adoption of such models can be considered. Two readers validated a previously published and well-performing logistic regression model on an external cohort. The model performance was not generalisable and offered no advantage to using PSAd cut-offs, and there was marked variation in model score related to contour differences from different readers. This potential variability should be investigated in future models which use quantitative MRI.

Xianling Qian¹, Changwu Zhou¹, Yunfei Zhang², Yongming Dai², and Mengsu Zeng^1
1Zhongshan Hospital, Fudan University, Shanghai, China, ²MR Collaboration, Central Research Institute, United Imaging Healthcare, Shanghai, China

Keywords: Liver, Radiomics

In the past decades, the incidence and mortality of intrahepatic cholangiocarcinoma (ICC) have been on the rise. A nuclear antigen named Ki67 is a poor prognostic predictor and an attractive therapeutic target in patients with ICC. Therefore, accurate prediction of Ki67 status in ICC patients is a predictor for treatment efficacy evaluation and outcome prediction. In this study, we established a multiparametric model for predicting Ki67 status in ICC patients preoperatively.

View Presentation Video

Changyu Sun^1,2, Craig A. Emter³, Darla L. Tharp³, and Talissa A. Altes^2
1Biomedical, Biological and Chemical Engineering, University of Missouri Columbia, Columbia, MO, United States, ²Radiology, University of Missouri Columbia, Columbia, MO, United States, ³Biomedical Sciences, University of Missouri Columbia, Columbia, MO, United States

Keywords: Lung, Machine Learning/Artificial Intelligence, Self-supervised Denoising

Self-supervised learning denoising networks can be applied to noisy only datasets when the clean-noisy pairs are not available, which is suitable for dynamic contrast-enhanced (DCE) pulmonary imaging where SNR is low and no ground truth clean image can be acquired. Blind-spot network with asymmetric pixel-shuffle downsampling (AP-BSN) was trained to utilize the advantages of self-supervised BSN and improve the denoising performance for pixel-wise independent and dependent noise. AP-BSN denoised images showed improved image quality by human reader assessment. AP-BSN showed generalization ability from human subjects to swine and from 1.5T to 3T.

Qian Wang¹, Qiu Bi², Yijun Zheng¹, Yaoxing Wang¹, Yuhui Chen², Chenrong Li¹, Xianhong Wang¹, Yunzhu Wu³, and Guoli Bi^2
1Medical school, Kunming University of Science and Technology, Kunming, China, ²Department of MRI, the First People’s Hospital of Yunnan Province, Kunming, China, ³MR Scientific Marketing, Siemens Healthineers Ltd, Shanghai, China

Keywords: Uterus, Radiomics, High-resolution MRI

Uterine prolapse has become one of the most common chronic diseases affecting women's health and quality of life. Early detection and intervention would be important factors to delay the progress of uterine prolapse. In this study, high-resolution magnetic resonance imaging(HR-MRI) of pelvic floor combined with radiomics was used to investigate the degree of uterine prolapse. It was found that the HR-MRI radiomics can be used as an effective evaluation method, and the nomogram has the best diagnostic efficacy for predicting the degree of uterine prolapse.

View Presentation Video

Nan Meng¹, Pengyang Feng², Xuan Yu¹, Yaping Wu¹, Fangfang Fu¹, Ziqiang Li³, Yu Luo¹, Hongna Tan¹, Jianmin Yuan⁴, Yang Yang⁵, Zhe Wang⁴, and Meiyun Wang*^6
1Department of Medical Imaging, Zhengzhou University People’s Hospital & Henan Provincial People’s Hospital, Zhengzhou, China, ²Department of Medical Imaging, Henan University People’s Hospital & Henan Provincial People’s Hospital, Zhengzhou, China, ³Department of Medical Imaging, Xinxiang Medical University Henan Provincial People’s Hospital, Zhengzhou, China, ⁴Central Research Institute, UIH Group, Shanghai, China, ⁵Beijing United Imaging Research Institute of Intelligent Imaging, UIH Group, Beijing, China, ⁶Zhengzhou University People’s Hospital & Henan Provincial People’s Hospital, Zhengzhou, China

Keywords: Cancer, PET/MR

¹⁸F-FDG PET/MRI is one of the most advanced means for the noninvasive diagnosis and evaluation of tumors. Three-dimensional ultrashort echo time (3D-UTE) is a novel MRI technique, which not only does not have ionizing radiation but also yields similar diagnostic results as conventional pulmonary CT. Our results showed that the ¹⁸F-FDG PET/MRI model based on clinical factors, 3D-UTE, and PET radiomics features could noninvasively assess the lymph node status in non-small cell lung cancer(NSCLC). Compared with the PET/CT model, it has similar diagnostic efficiency but less radiation, which is expected to provide new ideas for related research.

View Presentation Video

Mengsi Li¹, Keni Zheng², Ziying Yin², Lina Zhang¹, Jinhui Zhou¹, Dingyue Zhang¹, Jun Chen², Kevin J. Glaser², Richard L. Ehman², and Jin Wang^1
1The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China, ²Mayo Clinic, Rochester, MN, United States

Keywords: Cancer, Elastography, Hepatocellular carcinoma

A histologic sign known as “vessels encapsulating tumor clusters” (VETC) has been shown to be a powerful predictor of aggressive hepatocellular carcinoma (HCC) and is associated with unfavorable prognosis. It has been previously demonstrated that MR elastography (MRE)-based stiffness and shear strain mapping are promising in prediction of HCC aggressiveness. We investigated the diagnostic performance of MRE for predicting the VETC finding in HCC. Our results showed that 3D MRE-based peritumor OSS-pLSL and tumor stiffness performed well in predicting VETC status preoperatively, and their combination achieved an AUC of 0.92 in predicting VETC with sensitivity (87.9%) and specificity (83.9%).