CRCaroline Reinhold

Caroline Reinhold

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Caroline Reinhold has a total of 329 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Optimizing the MRI Pelvis Protocol for O-RADS MRI

    by Angela Tong, Nancy Kim, Krupa Patel-Lippmann, Stephanie Nougaret, Caroline Reinhold, Andrea G. Rockall, Isabelle Thomassin-Naggara, Elizabeth A. Sadowski

    Optimizing technical parameters to the specifications of the Ovarian-Adnexal Reporting and Data System (O-RADS) MRI protocol is necessary to achieve the positive predictive values listed in the O-RADS MRI risk table and to accurately assess the risk of cancer in lesions referred for MRI.


  • Physics Insights into Liver MRI: Educational Guidance for Protocol Optimization.

    by Jérémy Dana, Evan McNabb, Véronique Fortier, Khaled Alanazi, Akiyo Takada, Alice Fung, Amir A Borhani, Claude B Sirlin, An Tang, Caroline Reinhold, et al.

    MRI is the modality of choice for detection, characterization, and noninvasive diagnosis of focal liver observations. It offers excellent tissue contrast using multiple sequences and postcontrast multiphasic imaging. However, its diagnostic accuracy depends on appropriate selection of pulse sequences and imaging parameters. Poorly acquired images or suboptimal sequence parameter selection can lead to medical errors, such as missed lesions and inaccurate characterization of focal liver observations. A solid understanding of applied MRI physics can help optimize image quality and prevent such errors. In this educational review, the authors aim to (a) guide radiologists in assessing MRI quality, (b) promote effective communication using appropriate terminology when collaborating with technologists and medical physicists, and (c) establish a reference liver MRI protocol. First, they detail metrics of MRI quality-contrast-to-noise ratio, spatial resolution, and signal-to-noise ratio-along with the impact of key technical parameters on acquisition time and step-by-step guidance for improving image quality. Second, they review the range of pulse sequences used in liver MRI, discussing their respective strengths and limitations in achieving optimal imaging quality. Third, they address key ancillary considerations for performance of liver MRI, including fat suppression techniques, respiratory motion suppression strategies, and acceleration methods. ©RSNA, 2026 Supplemental material is available for this article.

  • Open Access

    Robustness of Deep Learning Segmentation Across Heterogeneous Uterine MRI Datasets: A Multi-Task Comparison of U-Net, Swin-UNETR, MedNeXt, and nnU-Net

    by Daniel A. Di Giovanni, Akiyo Takada, Evan McNabb, Jérémy Dana, Hajime Yokota, Takahiro Tsuboyama, Rita Zakarian, Martin Vallières, James Man Git Tsui, Caroline Reinhold

    Reliable medical image segmentation requires performance that is stable across tasks, acquisition settings, and evaluation domains. We evaluated U-Net, Swin-UNETR, and MedNeXt as manually configured architectures and nnU-Net as a self-configuring reference across three T2-weighted uterine MRI datasets: public multiclass uterine anatomy and fibroid segmentation (UMD, n=300), institutional endometrial cancer tumor segmentation (n=206), and institutional uterine mass lesion segmentation (n=234). A separate UMD-style cohort (n=12), relabeled to the same annotation ontology, was used to assess external domain shift. Fixed data partitions, fold ensembling, and matched evaluation were used across models. Performance was assessed with Dice, 95th-percentile Hausdorff distance, average symmetric surface distance, absolute volume difference, and paired bootstrap comparisons with within-dataset Holm correction. MedNeXt was the strongest manually configured architecture, whereas nnU-Net achieved the highest Dice on all internal datasets: 0.761 for UMD, 0.746 for endometrial cancer, and 0.814 for uterine mass segmentation, compared with 0.722, 0.726, and 0.789 for MedNeXt. The nnU-Net-MedNeXt separation was largest for multiclass UMD segmentation and smaller for the binary tasks. All models degraded substantially on external testing; Dice was 0.542 for nnU-Net, 0.490 for MedNeXt, 0.396 for U-Net, and 0.287 for Swin-UNETR. These findings show that architecture, task definition, and dataset-adaptive pipeline configuration jointly influence uterine MRI segmentation, while external domain shift remains a major source of failure.

  • Open Access

    Deep learning reconstruction in biparametric prostate MRI: Impact on qualitative and radiomics analyses

    by Jérémy Dana, Evan McNabb, Juan Castro, Ibtisam Al-Qanoobi, Yoshie Omiya, Kenny Ah-Lan, Véronique Fortier, Giovanni Artho, Caroline Reinhold, Simon Gauvin


  • Open Access

    CAR/CSAR Practice Statement on Pelvic MRI for Endometriosis

    by Emily Pang, Arvind Shergill, Silvia Chang, Priscila Crivellaro, Shauna Duigenan, Courtney R. Green, Candyce Hamel, Signy Holmes, Ania Kielar, Iffat Rehman, et al.

    The Canadian Association of Radiologists (CAR) Endometriosis Working Group was tasked with providing guidance and benchmarks to ensure the quality of technique and interpretation for advanced imaging modalities associated with diagnosing endometriosis. Advanced pelvic ultrasound is essential in diagnosing and mapping pelvic endometriosis, but pelvic MRI serves as an excellent imaging tool in instances where access to advanced ultrasound is limited, or an alternative imaging modality is required. Despite the known utility of MRI for endometriosis, there is no consensus on imaging protocol and patient preparation in Canada. To improve patient care and support excellence in imaging, the Working Group has developed recommendations for the use of pelvic MRI to assess for endometriosis with an aim to standardize MRI technique for use in both community and academic practices across Canada. The guidelines provide recommendations regarding imaging technique and patient preparation for pelvic MRI, along with suggestions for structured reporting of pelvic MRI for endometriosis.


  • Open Access

    CT-based deep learning prediction of complete response in intermediate-stage hepatocellular carcinoma treated with drug-eluting beads transarterial chemoembolization

    by Jérémy Dana, Armine Vardazaryan, Benoit Gallix, Maxime Ronot, Jean-Paul Mazellier, Marlee Parsons, Valérie Vilgrain, Thomas F Baumert, Caroline Reinhold, Nicolas Padoy, et al.

    Objectives: To develop a CT-based deep learning (DL) model to predict complete response (CR) to drug-eluting beads-transarterial chemoembolization (TACE) in patients with naive Barcelona Clinic Liver Cancer (BCLC) B hepatocellular carcinoma (HCC). Methods: This dual-centre retrospective study included 93 patients with BCLC B HCC treated with drug-eluting beads-TACE (median size of 40 mm and 37 mm at Institutions 1 and 2). Complete response was defined as per modified Response Evaluation Criteria in Solid Tumours on liver contrast-enhanced CT within 2 months of treatment. A twin-network DL model with spatio-temporal Video Vision Transformer (ViViT) architecture was developed to predict CR using baseline dedicated liver CT. The model was extensively trained/tested based on an 8-fold cross-validation approach with an ensemble technique, a model vote system where the outcome is the average of multiple model predictions. Results: The CR rate was 36% (18/50) and 22% (11/49) at Institutions 1 and 2. The model showed high specificity and AUC, as well as moderate sensitivity and balanced accuracy when using either the late arterial phase (0.91 ± 0.12, 0.86 ± 0.16, 0.43 ± 0.23, and 0.67 ± 0.13, respectively) or the portal venous phase (0.90 ± 0.15, 0.85 ± 0.17, 0.57 ± 0.30, and 0.74 ± 0.16, respectively). Conclusion: The developed CT-based DL model predicted CR to drug-eluting beads-TACE in patients with naive BCLC B HCC with high specificity. It should be refined to improve sensitivity. Advances in knowledge: The study underscores the potential of artificial intelligence in precision medicine in patients with HCC. While the model shows promise, further research with larger datasets and prospective studies is needed to enhance its predictive power and clinical applicability.


  • MRI of the Rectum: A Decade into DISTANCE, Moving to DISTANCED

    by Stephanie Nougaret, Kirsten Gormly, Doenja M.J. Lambregts, Caroline Reinhold, Vicky Goh, Elena Korngold, Quentin Denost, Gina Brown

    Radiologists play a vital role in rectal cancer decision-making, and updating the DISTANCE mnemonic to the MRI-based DISTANCED offers comprehensive diagnostic insights and emphasizes its importance for informed decision-making for patients.


  • Open Access

    R2* Impact on Hepatic Fat Quantification With a Commercial Single Voxel Technique at 1.5 and 3.0 T

    by Véronique Fortier, Ahmed Mohamed, Evan McNabb, Jérémy Dana, Rita Zakarian, Ives R. Levesque, Caroline Reinhold

    Rationale and Objectives: Fat quantification accuracy using a commercial single-voxel high speed T2-corrected multi-echo (HISTO) technique and its robustness to R2* variations at 3.0 T, such as those introduced by iron in liver, has not been fully established. This study evaluated HISTO at 3.0 T and sought to reproduce results at 1.5 T. Methods: Phantoms were prepared with a range of fat content and R2*. Data were acquired at 1.5 T and 3.0 T, using HISTO and a Dixon technique. Fat quantification accuracy was evaluated as a function of R2*. The patient study included 239 consecutive patients. Data were acquired at 1.5 T or 3.0 T, using HISTO and Dixon techniques. The techniques were compared using Bland-Altman plots. Bias significance was evaluated using a one-sample t-test. Results: In phantoms, HISTO was accurate within 10% up to a R2* of 100 s−1 at both field strengths, while Dixon was accurate within 10% where R2* was accurately quantified (up to 350 s−1 at 1.5 T, and 550 s−1 at 3.0 T). In patients, where R2* was −1, fat quantification from both techniques agreed at 1.5 T ( P = .71), but not at 3.0 T ( P = .007), with a bias 2* is 100 s−1. Dixon should be preferred for hepatic fat quantification due to its robustness to R2* variations.


  • Top Cited Papers

    O-RADS US v2022: An Update from the American College of Radiology’s Ovarian-Adnexal Reporting and Data System US Committee

    by Lori M. Strachowski, Priyanka Jha, Catherine H. Phillips, Misty M. Blanchette Porter, Wouter Froyman, Phyllis Glanc, Yang Guo, Maitray D. Patel, Caroline Reinhold, Elizabeth J. Suh-Burgmann, et al.

    First published in 2019, the Ovarian-Adnexal Reporting and Data System (O-RADS) US provides a standardized lexicon for ovarian and adnexal lesions, enables stratification of these lesions with use of a numeric score based on morphologic features to indicate the risk of malignancy, and offers management guidance. This risk stratification system has subsequently been validated in retrospective studies and has yielded good interreader concordance, even with users of different levels of expertise. As use of the system increased, it was recognized that an update was needed to address certain clinical challenges, clarify recommendations, and incorporate emerging data from validation studies. Additional morphologic features that favor benignity, such as the bilocular feature for cysts without solid components and shadowing for solid lesions with smooth contours, were added to O-RADS US for optimal risk-appropriate scoring. As O-RADS US 4 has been shown to be an appropriate cutoff for malignancy, it is now recommended that lower-risk O-RADS US 3 lesions be followed with US if not excised. For solid lesions and cystic lesions with solid components, further characterization with MRI is now emphasized as a supplemental evaluation method, as MRI may provide higher specificity. This statement summarizes the updates to the governing concepts, lexicon terminology and assessment categories, and management recommendations found in the 2022 version of O-RADS US.


  • Adnexal Lesion Imaging: Past, Present, and Future

    by Elizabeth A. Sadowski, Andrea Rockall, Isabelle Thomassin-Naggara, Lisa M. Barroilhet, Sumer K. Wallace, Priyanka Jha, Akshya Gupta, Atul B. Shinagare, Yang Guo, Caroline Reinhold

    Currently, imaging is part of the standard of care for patients with adnexal lesions prior to definitive management. Imaging can identify a physiologic finding or classic benign lesion that can be followed up conservatively. When one of these entities is not present, imaging is used to determine the probability of ovarian cancer prior to surgical consultation. Since the inclusion of imaging in the evaluation of adnexal lesions in the 1970s, the rate of surgery for benign lesions has decreased. More recently, data-driven Ovarian-Adnexal Reporting and Data System (O-RADS) scoring systems for US and MRI with standardized lexicons have been developed to allow for assignment of a cancer risk score, with the goal of further decreasing unnecessary interventions while expediting the care of patients with ovarian cancer. US is used as the initial modality for the assessment of adnexal lesions, while MRI is used when there is a clinical need for increased specificity and positive predictive value for the diagnosis of cancer. This article will review how the treatment of adnexal lesions has changed due to imaging over the decades; the current data supporting the use of US, CT, and MRI to determine the likelihood of cancer; and future directions of adnexal imaging for the early detection of ovarian cancer.