Martins Otikovs¹, Noam Nissan², Edna Furman-Haran¹, Debbie Anaby², Tanir M. Allweis³, Ravit Agassi⁴, Miri Sklair-Levy^2,5, and Lucio Frydman^1
1Weizmann Institute of Science, Rehovot, Israel, ²Sheba Medical Center, Ramat Gan, Israel, ³Kaplan Medical Center, Rehovot, Israel, ⁴Ben Gurion University Hospital, Beer Sheba, Israel, ⁵Tel Aviv University, Tel Aviv, Israel

Spatiotemporal encoding (SPEN) is an alternative ultrafast imaging technique which allows to overcome distortions otherwise observed along EPI’s phase-encoded dimension, and to perform self-referenced phase corrections in interleaved diffusion weighted imaging (DWI) scans. This study compares SPEN’s performance against multishot read-out segmented EPI (RESOLVE), with an emphasis on high-b-valued DWI and diffusion kurtosis imaging in the context of breast cancer imaging. The results show SPEN’s advantages for delivering kurtosis maps that provide a separation between cancerous and healthy tissues. The potential of using SPEN DW images acquired with high b-values as an alternative to DCE subtractions is also assessed.

Debosmita Biswas¹, Justin Romanoff², Dariya Malyarenko³, Wesley Surrento⁴, Habib Rahbar¹, Nola Hylton⁵, David C Newitt⁵, Thomas L Chenevert³, and Savannah C Partridge^1
1Radiology, University of Washington, Seattle, WA, United States, ²Center for Statistical Sciences, Brown University, Providence, RI, United States, ³Radiology, University of Michigan, Ann Arbor, MI, United States, ⁴Biomedical and Health Informatics, University of Washington, Seattle, WA, United States, ⁵Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, United States

Gradient nonlinearity (GNL) correction shows potential to improve the accuracy of ADC values collected across different MRI platforms. Here, we retrospectively applied GNL correction to breast DWI datasets collected in the ECOG-ACRIN A6702 trial by pixel-wise scaling of the ADC map with correction factor map. Our findings confirm that GNL significantly impacts multicenter breast lesion ADC values, and that GNL-based ADC errors vary significantly across MRI vendors and gradient systems. Therefore, GNL correction is important for implementation of generalizable ADC thresholds for separating benign and malignant lesions. 

Natsuko Onishi¹, Jessica Gibbs¹, Teffany Joy Bareng¹, Wen Li¹, Elissa R. Price¹, Bonnie N. Joe¹, Laura J. Esserman², The I-SPY 2 Consortium³, David C. Newitt¹, and Nola M. Hylton^1
1Department of Radiology & Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, ²Department of Surgery, University of California, San Francisco, San Francisco, CA, United States, ³Quantum Leap Healthcare Collaborative, San Francisco, CA, United States

In the I-SPY2 neoadjuvant breast cancer trial, functional tumor volume (FTV) derived from dynamic contrast-enhanced MRI serves as a key marker. Participants in I-SPY2 have the option to “de-escalate” therapy if achievement of pathological complete response (pCR) is highly likely at inter-regimen time point. A model combining FTV-based predictive probabilities with inter-regimen core-biopsy pathology is central to select candidates for this option. This retrospective study compared the performance of longitudinal FTVs in predicting pCR between optimal and non-optimal FTV segmentation groups. The results suggest that improvements to FTV segmentation can improve FTV’s ability to provide predictive guidance for treatment de-escalation.

Rie Ota¹, Masako Kataoka¹, Maya Honda¹, Mami Iima¹, Kanae Kwai Miyake¹, Akane Ohashi², Yosuke Yamada³, Masakazu Toi⁴, and Yuji Nakamoto^1
1Department of Diagnostic Imaging and Nuclear Medicine, Kyoto University graduate school of medicine, Kyoto, Japan, ²Kyoto Medical Center, Kyoto, Japan, ³Department of Pathology, Kyoto University Hospital, Kyoto, Japan, ⁴Department of Breast Surgery, Kyoto University Hospital, Kyoto, Japan

This study aimed to examine the performance of DWI in diagnosing pCR before surgery. Kinetic analysis from standard DCE-MRI were analyzed for comparison. ROC analysis for diagnosing pCR based on DWI score/Kinetic score by two readers was performed. Kinetic score showed slightly higher AUC while 95% confidence interval overlapped with that of DWI score. Both kinetic score and DWI score demonstrated excellent diagnostic performance among triple negative subtype compared to other subtypes.

Bonny Chau¹, Debosmita Biswas¹, Anum S. Kazerouni¹, Daniel S. Hippe¹, Rebeca Alvarez¹, Suzanne Dintzis¹, Laura C. Kennedy², Vijayakrishna Gadi³, and Savannah C. Partridge ^1
1University of Washington, Seattle, WA, United States, ²Vanderbilt University, Nashville, TN, United States, ³University of Illinois, Chicago, IL, United States

We investigated the relationship between immune infiltration and imaging metrics derived from breast MRI in patients with HER2+ breast cancer treated with trastuzumab. Fourteen patients with localized HER2+ breast cancer were imaged with diffusion-weighted and dynamic contrast-enhanced (DCE-) MRI prior to and ~2 weeks after a run-in dose of trastuzumab. Pre-treatment ADC and change in DCE-MRI peak percent enhancement were both significantly associated with immune response as characterized by change in level of tumor infiltrating lymphocytes.

Jiejie Zhou¹, Yan-Lin Liu², Yang Zhang², Jeon-Hor Chen^2,3, Freddie J. Combs², Ritesh Parajuli⁴, Rita S. Mehta⁴, Huiru Liu¹, Zhongwei Chen¹, Youfan Zhao¹, Meihao Wang¹, and Min-Ying Su^2
1Department of Radiology, First Affiliated Hospital of Wenzhou Medical University, Wenzhou, China, ²Department of Radiological Sciences, University of California, Irvine, CA, United States, ³Department of Radiology, E-Da Hospital and I-Shou University, Kaohsiung, Taiwan, ⁴Department of Medicine, University of California, Irvine, CA, United States

A total of 150 lesions, 104 malignant and 46 benign, presenting as non-mass-like enhancements were analyzed. Three radiologists performed BI-RADS reading for the morphological distribution and internal enhancement pattern. For each case, the 3D tumor mask was generated using Fuzzy-C-Means segmentation. Three DCE parametric maps were generated, and PyRadiomics was applied to extract features. The radiomics model was built using 5 different machine learning algorithms. ResNet50 was implemented using three parametric maps as input. SVM yielded the highest accuracy of 80.4% in training, 77.5% in testing datasets. ResNet50 had better diagnostic performance, 91.5% in training, and 83.3% in testing datasets.

Jing Zhang¹, Chenao Zhan², Tao Ai², Xu Yan³, and Guang Yang^1
1Shanghai key lab of magnetic resonance, shanghai, China, ²Tongji Medical College, Huazhong University of Science and Technology, Department of Radiology，Tongji Hospital, Wuhan, Hubei Province, China, ³Siemens Healthcare, MR Scientific Marketing, shanghai, China

    DCE is the most useful MRI sequence for breast cancer diagnosis, but it often suffers from a high rate of false positive. To overcome this problem, we combined radiomics features from DCE, T2W, and DWI images to build a new machine learning model for differentiation of breast cancer. Our model achieved an AUC of 0.948 in an internal test cohort and 0.944 in an external test cohort, and reduced the false positive rate effectively. It was also found, first-order and texture features from ADC map made significant contributions to the model, suggesting the value ADC in breast cancer classification.   

Dipika Sikka^1,2, Nanyan Zhu³, Chen Liu⁴, Scott Small⁵, and Jia Guo^6
1Department of Biomedical Engineering, Columbia University, New York, NY, United States, ²VantAI, New York, NY, United States, ³Department of Biological Sciences and the Taub Institute, Columbia University, New York, NY, United States, ⁴Department of Electrical Engineering and the Taub Institute, Columbia University, New York, NY, United States, ⁵Department of Neurology, the Taub Institute, the Sergievsky Center, Radiology and Psychiatry, Columbia University, New York, NY, United States, ⁶Department of Psychiatry, Mortimer B. Zuckerman Mind Brain Behavior Institute, Columbia University, New York, NY, United States

Gadolinium-based contrast agents (GBCAs) have facilitated an improved analysis and understanding of structural lesions, however, present safety risks due to the tissue retention of GBCAs. Here we optimize and apply the deep learning model, DeepContrast, to predict gadolinium uptake in brain and breast structural lesions for structural lesion enhancement. The optimized DeepContrast models predict gadolinium uptake that is comparable to ground-truth scans consisting of the uptake from the GBCAs, using a single T1-weighted pre-contrast scan.

Bin Deng^1,2,3, Mansi Saksena^2,3, Steven Jay Isakoff^3,4, Ralph Sinkus⁵, Samuel Patz^3,6, and Stefan Alexandru Carp^1,2,3
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, ²Department of Radiology, Massachusetts General Hospital, Boston, MA, United States, ³Harvard Medical School, Boston, MA, United States, ⁴Cancer Center, Massachusetts General Hospital, Boston, MA, United States, ⁵Laboratory for Vascular Translational Science (LVTS), Institut National de la Santé et de la Recherche Médicale (INSERM), Paris, France, ⁶Department of Radiology, Brigham and Women’s Hospital, Boston, MA, United States

Breast cancers are complex, evolving systems characterized by profound spatial and temporal heterogeneity in their biological nature. A multimodal multiparametric approach is needed to synergistically use imaging methods with different biophysical basis to simultaneously quantify multiple aspects of tumor physiology. We built a custom breast coil to allow multimodal near-infrared diffuse optical tomography and MR elastography imaging of human breast. Results of a three-inclusion dual-contrast phantom showed clear contrasts in reconstructed mechanical and optical properties as expected. Data on a breast cancer patient showed collocated hemoglobin and stiffness contrast at the tumor location.

Nabil Elshafeey¹, Gaiane M. Rauch², Aikaterini Kotrotsou³, Beatriz E. Adrada¹, Rosalind P. Candelaria¹, Abeer H. Abdelhafez¹, Huiqin Chen⁴, Jia Sun⁴, Medine Boge¹, Rania M. M Mohamed¹, Benjamin C. Musall⁵, Jong Bum Son⁵, Shu Zhang⁶, Jason B. White⁷, Brandy Willis⁵, Elizabeth Ravenberg⁷, Wei Peng⁴, Stacy L. Moulder⁷, Wei Yang¹, Mark D. Pagel⁶, Jingfei Ma⁵, and Ken-Pin Hwang^5
1Breast Imaging, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ²Breast and Abdominal imaging, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ³The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ⁴Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ⁵Imaging Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ⁶Cancer Systems Imaging, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ⁷Breast Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, United States

Early identification of treatment response to neoadjuvant systemic therapy (NAST) in Triple Negative Breast Cancer (TNBC) patients is important for appropriate treatment selection and response monitoring. In this study we evaluated the ability of a radiomic model extracted from a novel sequence, Magnetic Resonance Image Compilation (MAGIC), acquired before treatment initiation, to predict NAST response in TNBC. Our results showed that the radiomic signature derived from MAGIC maps (T1, PD and T2) can help differentiate responders from non-responders at baseline evaluation.