View Presentation Video

Keywords: Radiomics, Machine Learning/Artificial Intelligence, Data Modelling

Radiomics uses quantitative analysis of medical imaging based on machine learning techniques and has shown its potentials of aiding personalized clinical decisions. A high standard of clinical reference (or ground truth, endpoint) is vital in radiomics feature selection and modeling, but is commonly overlooked, and assumed to be perfect. However, in reality, there are uncertainties and variability in these clinical references due to many factors. We aim to quantitatively assess the influence of clinical reference uncertainty and variability on MRI Radiomics modeling via endpoint annotation permutation with different levels.

View Presentation Video

Jixin Luan¹, Chuanchen Zhang², Di Zhang², and Guolin Ma^1
1China-Japan Friendship Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China, ²Department of Radiology, Liaocheng People's Hospital, Shandong First Medical University & Shandong Academy of Medical Sciences, Liaocheng, China

Keywords: Radiomics, Brain, GBM

In this study, we aimed to investigate the correlation between immune-related lncRNA signature and radiomics signature and immune cell infiltration and immune checkpoint blockade in glioblastoma multiform (GBM), and to develop a prognostic model to predict the overall survival of patients. We found that immune-related lncRNA signature and radiomics signature can better predict the immune status and overall survival of GBM patients, which can help clinical prognosis determination and immunotherapy selection.

View Presentation Video

Harri Merisaari¹, Janne Verho², Ileana Montoya Perez^2,3, Otto Ettala⁴, Kari T Syvänen⁴, Pekka Taimen⁵, Aida Steiner², Jani Saunavaara⁶, Ekaterina Saukko², Peter Boström⁴, Hannu Aronen¹, and Ivan Jambor^1,2
1Department of Radiology, University of Turku, Raisio, Finland, ²Department of Radiology, Turku University Hospital, Turku, Finland, ³Department of Computing, University of Turku, Turku, Finland, ⁴Department of Urology, TYKS Turku University Hospital and University of Turku, Turku, Finland, ⁵Department of Pathology, TYKS Turku University Hospital, Turku, Finland, ⁶Department of Medical Physics, TYKS Turku University Hospital, Turku, Finland

Keywords: Radiomics, Cancer, Inter-site reproducibility, Deep Learning, bi-parametric MRI

In the current study, we aimed to explore reproducibility of various hand-crafted radiomics features, and deep learning autoencoder-based features within Gleason Grade Groups (GGG) using MULTI-IMPROD trial data. Differences between sites were evaluated with ANOVA test, corrected for GGG group, and multi-class AUC for GGG. We explored if systematic differences exist between the four centers taking part in the trial with conventionally used radiomic features. The results show differences between modalities, feature groups, and when intensity harmonization is applied for ADC. 

No Video Available

View Presentation Video

Pao-Han Chiu¹, Ming-Ting Wu², Ken-Pen Weng^3,4, Nai-Yu Pan⁵, Teng-Yi Huang⁵, and Hsu-Hsia Peng^1
1Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Hsinchu, Taiwan, ²Department of Radiology, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ³Department of Pediatrics, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ⁴Department of Pediatrics, National Yang Ming Chiao Tung University, Taipei, Taiwan, ⁵Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan

Keywords: Radiomics, Radiomics

Myocardial motion influences the stability of radiomic features of cardiac magnetic resonance (CMR) cine images. We aimed to evaluate the impact of the stability of myocardial radiomic features on the classification performance of differentiation of repaired tetralogy of Fallot (rTOF) patients from normal volunteers. The stability of each radiomic feature during cardiac cycle was assessed by coefficient of variation (CV) of the feature value. 50 of 107 radiomic features (46.7%) in normal volunteers were stable features. The classification model established only with stable features presented the best classification performance (AUC=0.94) which was able to be improved by chi-square feature selection.

Mehri Mehrnia¹ and Mohammed S.M. Elbaz^1
1Northwestern University, Chicago, IL, United States

Keywords: Radiomics, Cardiomyopathy, heart, cardiovascular, LGE, scar, Myocardial Infarction

There remains no standard method for quantification of myocardial infarction (MI) from LGE images. Current methods lack reproducibility as they rely on manual or threshold-based scar segmentation. Notably, these methods only quantify scar volume/percentage but ignore the impact of myocardial LGE distribution pattern. Here, we propose a novel threshold-free concept that comprehensively quantifies scar burden in terms of both scar extent and the unique myocardial LGE distribution pattern: LGE Scar Burden Signatures. We demonstrated our technique’s strong correlation to scar extent, its independent association with serum biomarker of myocardial injury (Troponin), and reduced ejection fraction, independent of Scar percentage.

Shiqi Chen¹, Yawen Xiao¹, Zhaotao Zhang¹, Jiankun Dai², Yifei Gui¹, and Xinlan Xiao^1
1Department of Radiology, The Second Affiliated Hospital of Nanchang University, Nanchang, China, ²GE Healthcare, MR Research China, Beijing, China

Keywords: Radiomics, Epilepsy

Focal cortical dysplasia (FCD) is the most common epileptogenic developmental malformation which remains challenging to diagnose. This study aimed to investigate the possibility of diagnosing FCD and laterality in epilepsy patients using the multiparametric MRI radiomics model. Radiomic features were extracted from the preoperative MRI images of 86 patients. Multivariable logistic regression was used to develop the diagnosis model. The performance of radiomic model was evaluated with general evaluation metrics and was compared with that of inexperienced radiologists.We concluded that radiomics features derived from the combined of T1WI, T2WI and FLAIR might help diagnosing FCD and laterality.

Wenjie Tang¹, Yuan Guo¹, Siyi Chen¹, Bingsheng Huang², Xiaotong Xie², Mingyu Wang², Yongzhou Xu³, Kuiming Jiang⁴, and Xinhua Wei^1
1Guangzhou First People's Hospital, Guangzhou, China, ²Shenzhen University, Shenzhen, China, ³Philips Healthcare, Guangzhou, China, ⁴Guangdong Women and Children Hospital, Guangzhou, China

Keywords: Radiomics, Breast, HER2 expressing; MRI; Deep learning; Prognosis

To the best of our knowledge, our study is the first to non-invasively assess human epidermal growth factor receptor 2 (HER2) status, especially HER2-low-positive status in breast cancer. In this study, a deep learning radiomics (DLR) model based on contrast-enhanced MRI was constructed and showed high and stable performance in predicting HER2 status in both the training and validation cohorts, and the predicted status was an independently significant predictor of disease-free survival (DFS) in HER2-low-positive/HER2-zero breast cancers. The DLR model showed prospects as a computer-aided diagnostic tool to help more accurately identify HER2-low-positive breast cancers, thereby guiding patient treatment strategies.

View Presentation Video

Keywords: Radiomics, Alzheimer's Disease

Alzheimer's disease causes significant gray matter loss, which leads to changes in the brain surface's shape. In this work, we used the local gyrification index (LGI) and the three-dimensional tortuosity (𝜏3𝐷) to characterize the cortical morphology, and to determine if the obtained values were significantly different amog the Alzheimer's diagnosis. The subset of MRI studies was obtained from ADNI database. For the data analyzed, the results show that the 𝜏3𝐷 has a positive correlation with brain volume and can potentially be a biomarker for AD.

View Presentation Video

Gaiying Li¹, Qifan Pang¹, Mengying Chen¹, Yang Song², Qing Cai³, Kai Zhang³, Longnian Lin³, Yi Wang⁴, and Jianqi Li^1,3
1School of Physics and Electronic Science, East China Normal University, Shanghai, China, ²MR Scientific Marketing, Siemens, Siemens Healthineers, Shanghai, China, ³Institute of Brain and Education Innovation, East China Normal University, Shanghai, China, ⁴Department of Radiology, Weill Medical College of Cornell University, New York, NY, United States

Keywords: Data Analysis, Quantitative Susceptibility mapping

Tissue iron play a critical role in cognitive functions. However, associations of basal ganglia iron concentration with cognition and emotion in children are less well understood. This study examined the correlation of susceptibility values in the bilateral basal ganglia nuclei with cognitive functions and social emotional capacity in children around the age of seven. The results highlighted that the inhibitory control, collaboration, open-mindedness showed significant association with susceptibility values in the basal ganglia. In conclusion, this QSM study indicated the potential for using brain iron content in the basal ganglia to assess cognitive and emotional performance during the children development.

View Presentation Video

Ana Beatriz Solana^1,2, Sagar Mandava³, Xinzeng Wang⁴, Marc Lebel⁵, David J Lythgoe², Tobias C Wood², Matthew Bowdler², Steven CR Williams², and Florian Wiesinger^1,2
1GE Healthcare, Munich, Germany, ²King's College London, London, United Kingdom, ³GE Healthcare, Marietta, GA, United States, ⁴GE Healthcare, Houston, TX, United States, ⁵GE Healthcare, Calgary, AB, Canada

Keywords: Pulse Sequence Design, MR Value, Neuro, quiet, multicontrast, fMRI, QSM

Most of today’s MRI scans are very loud and can be challenging for patients of all ages, e.g. sleeping babies, anxious children or elderly with tinnitus.  Existing quiet and/or acoustic noise reduced MRI scanning solutions are in their infancy, primarily because of their incomplete nature and associated trade-offs in terms of image quality and/or scan time.  Here, we present a comprehensive quiet neuroimaging solution, including standard anatomical scans (localizer, T1w-MPRAGE, T2w-FSE, T2w-FLAIR, MRA, Diffusion), quantitative parameter mapping (ADC, T2*, QSM, PD, T1, T2) and functional BOLD-fMRI.  Scan time and overall image quality was improved using DL-based image reconstruction framework.

View Presentation Video

Keywords: Software Tools, Brain, Brainstem atlas, segmentation, functional and structural connectome, 7 Tesla MRI, 3 Tesla MRI

Brainstem nuclei are deep gray matter regions involved in vital functions such as arousal/sleep/motor/sensory/autonomic/nociceptive/limbic/sensory function. Due to their small size and limited MRI contrast, these nuclei are difficult to visualize in conventional imaging of living humans. We generated and released the Brainstem Navigator toolkit, which enables the automatic identification of brainstem nuclei location in conventional and advanced MRI of living humans. It includes in-vivo probabilistic atlas labels for 31 brainstem nuclei generated by multi-contrast 7 Tesla MRI. We also developed coregistration routines optimized for the brainstem in 3 Tesla and 7 Tesla MRI data in health and disease.

View Presentation Video

William Zaylor^1,2, Sameed Khan^1,2,3, Richard Lartey^1,2, Brendan L. Eck^2,3, Mei Li^1,2, Sibaji Gaj^1,2, Jeehun Kim^1,2, Carl S. Winalski^1,2,3, Faysal Altahawi^2,3, Morgan H. Jones⁴, Laura J. Huston⁵, Kevin D. Harkins⁶, Michael V. Knopp⁷, Christopher C. Kaeding⁸, Kurt P. Spindler^2,9, and Xiaojuan Li^1,2,3
1Biomedical Engineering, Cleveland Clinic, Cleveland, OH, United States, ²Program of Advanced Musculoskeletal Imaging (PAMI), Cleveland Clinic, Cleveland, OH, United States, ³Department of Diagnostic Radiology, Cleveland Clinic Imaging Institute, Cleveland, OH, United States, ⁴Department of Orthopaedic Surgery, Brigham and Women's Hospital, Boston, MA, United States, ⁵Department of Orthopaedics and Rehabilitation, Vanderbilt University, Nashville, TN, United States, ⁶Radiology and Radiological Sciences, Vanderbilt University, Nashville, TN, United States, ⁷Department of Radiology, The Ohio State University Wright Center of Innovation in Biomedical Imaging, Columbus, OH, United States, ⁸Department of Orthopaedic Surgery, The Ohio State University, Columbus, OH, United States, ⁹Department of Orthopaedic Surgery, Cleveland Clinic, Cleveland, OH, United States

Keywords: Radiomics, Relaxometry

The purpose of this study was to evaluate the associations between radiomic features in knee and thigh quantitative MRI with symptomatic and radiographic post-traumatic osteoarthritis (PTOA) after anterior cruciate ligament reconstruction (ACLR). Radiomic features of knee cartilage, menisci (extracted from T_1ρ maps), and thigh muscles (extracted from anatomic images and fat fraction maps) were extracted from patient MRI 10+ years after ACLR, and features were selected for symptomatic or radiographic PTOA associations. Symptomatic and radiographic PTOA were associated with features from the medial cartilage compartments, and menisci; features from the thigh muscles were associated with symptomatic PTOA only.

View Presentation Video

Keywords: Data Acquisition, Data Acquisition

Echo planar time-resolved imaging (EPTI) is a recently developed multi-shot EPI method that provides time-resolved multi-echo images and providing multi-contrast images. Based on this basis, here we present Multi-TR-EPTI, a technique developed to achieve distortion-free and Multi-parametric quantitative MRI including T1, T2, T2*, and PD mapping. Multi-TR-EPTI exploits an optimized spatiotemporal CAIPI encoding in the k-TE-TR space.which can recover the k-TE-TR space within the same shot number as the original EPTI. A subspace reconstruction was employed to obtain hundreds of high-quality multi-contrast images at sub-millisecond temporal increments. Multi-TR-EPTI adopts different TRs to provide T1 contrast while improving scan efficiency.

View Presentation Video

Lin WU¹, Chris Carchi², Shalom Michaeli¹, Silvia Mangia¹, and Djaudat Idiyatullin^1
1CMRR and Department of Radiology, University of Minnesota, Minneapolis, MN, United States, ²University of Minnesota, Minneapolis, MN, United States

Keywords: High-Field MRI, Relaxometry, T1ρ

A new method entitled alternating Look-Locker (aLL) for mapping of T₁ and T_1ρ is proposed. Magnetization preparation modules incorporating adiabatic full passage pulses were imbedded in a Look-Locker scheme that additionally alternates magnetization from +Z and -Z axes. MB-SWIFT was used as a readout. Analytical derivations and relevant simulations are presented. Phantom experiments and in vivo studies in the rat brain were conducted at 9.4 T. Results show that aLL allows more robust and faster acquisitions as compared to previously introduced steady–state technique, providing the possibility of simultaneous collection of T₁ and T_1ρ maps in one acquisition.