Research record

Publications, talks & posters

Deep learning applied to musculoskeletal, breast, neuro, thoracic and cardiac imaging, along with teaching radiologists how to read and build AI studies.

Case studies 38 publications 30 oral presentations 9 posters Leadership & honors CV (PDF)

Case studies

Four studies, in brief

Problem, contribution, method, result and limitation for four published studies. Numbers are taken from the final publications. All are research models, not deployed clinical products.

Breast imaging | Acad Radiol 2019 · Co-author; model developer

Breast cancer risk from the mammogram itself

Clinical problem: Breast density is a coarse risk marker. The question was whether a network reading the whole mammogram could stratify future risk better.
My contribution: I wrote the source code and programmed the CNN risk model, working in Columbia’s breast imaging AI group with Ha (first author) and Jambawalikar (senior author). This line of work led to my NIH StARR-funded CAPRI resident research year.
Approach: Retrospective case-control study: prior mammograms of 210 women who later developed cancer (420 images) vs. 527 controls (1,054 images). A fully convolutional network produced a pixel-wise risk map, with an 80/20 split and five-fold cross-validation.
Published finding: The network score and breast density were independent predictors. The network’s odds ratio was 4.42 (95% CI 3.4–5.7) vs. 1.67 (1.4–1.9) for density, with 72% accuracy in identifying the case group.
Limitation: Single institution, retrospective and relatively small. The authors expected more data to improve the model. A research model.
Read more: Paper (DOI) Full text (PMC) CAPRI award announcement

Multivariate odds ratios as published.

MSK · Pediatric | J Digit Imaging 2018 · First author

Machine-assisted bone age (MABAL)

Clinical problem: Bone-age reads against the Greulich–Pyle atlas are common, slow and observer-dependent. Earlier deep-learning attempts were limited by small datasets and generic architectures.
My contribution: First author, working with Columbia co-authors Chang, Ruzal-Shapiro and Ayyala.
Approach: 10,289 hand radiographs from our PACS and the public Digital Hand Atlas, split into four sex-and-age cohorts. A custom 14-layer network with residual, inception and spatial-transformer layers was trained from scratch, with augmentation to limit overfitting.
Published finding: Aggregate mean absolute error of 0.637 years on validation and 0.536 years on the test set, where the reference standard averaged two pediatric radiologists. At publication, the best reported deep-learning result for bone age.
Limitation: One institution’s data plus a public atlas, and training labels came from single clinical reads. A research model, not a clinical product.
Read more: Paper (DOI) Full text (PMC) Related code: RSNA bone-age challenge

Study design as described in the paper. Test MAE is the aggregate across the four cohorts.

MSK · Emergency | J Digit Imaging 2020 · First author

Femoral neck fracture detection and Garden classification

Clinical problem: Hip fractures need to be found and classified quickly because displacement drives the surgical choice, but labelled examples, especially nondisplaced Garden I/II fractures, are scarce.
My contribution: First author, working with Columbia MSK co-authors Varada, Goel, Wong and Rasiej.
Approach: 1,063 AP hip radiographs from 550 patients. One network localized the femoral neck and a second classified Garden I/II, Garden III/IV or no fracture. Training data were expanded to 9,063 images with GAN-generated radiographs and digitally reconstructed radiographs (DRRs) from preoperative hip CT.
Published finding: On 105 held-out real radiographs, fracture vs. no fracture reached AUC 0.92, accuracy 92.3%, sensitivity 0.91 and specificity 0.93. The three-class task reached AUC 0.96 and accuracy 86.0%.
Limitation: Retrospective, single-institution data with a small test set; only 127 real Garden I/II radiographs. A research model, not deployed clinically.
Read more: Paper (DOI) Full text (PMC) Health Imaging coverage

Two-class AUC on the held-out set, as reported for each augmentation strategy. Axis starts at zero.

AI evaluation · MSK | J Imaging Inform Med 2024 · Co-author; ground-truth reader

Are saliency maps trustworthy explanations?

Clinical problem: Heatmaps are widely used to show “where the AI looked,” but their reliability had not been tested on musculoskeletal radiographs.
My contribution: Co-author, and one of the three fellowship-trained MSK radiologists who drew the ground-truth boxes, on a Johns Hopkins and University of Maryland study led by Venkatesh (first author) and Yi (senior author).
Approach: Networks trained on the public Stanford MURA upper-extremity dataset. Six gradient-based saliency methods were scored on localization, repeatability, reproducibility and sensitivity to model weights, against consensus boxes from three fellowship-trained MSK radiologists.
Published finding: Every method localized abnormalities worse than radiologists (AUPRC 0.438–0.590 vs. 0.816) and was less repeatable and reproducible. Five passed the weight-sensitivity test, but none met all four criteria. The authors recommend caution before clinical use.
Limitation: One public dataset of upper-extremity radiographs, and gradient-based methods only. Results may differ for other explanation methods, body parts or modalities.
Read more: Paper (DOI) SIIM 2022 podium presentation

Reported ranges across the six methods versus the radiologist average. Higher is better.

Peer-reviewed

Journal publications

38 articles. Full author lists are available via each DOI.

2024

Effects of vitamin D supplementation on a deep learning-based mammographic evaluation in SWOG S0812.McGuinness JE, Anderson GL, Mutasa S, et al. JNCI Cancer Spectr. 2024. DOI for “Effects of vitamin D supplementation on a deep learning-based mammographic evaluation in SWOG S0812”
Gradient-based saliency maps are not trustworthy visual explanations of automated AI musculoskeletal diagnoses.Venkatesh K, Mutasa S, Moore F, Sulam J, Yi PH. J Imaging Inform Med. 2024. DOI for “Gradient-based saliency maps are not trustworthy visual explanations of automated AI musculoskeletal diagnoses”
Deep learning to detect left ventricular structural abnormalities in chest X-rays.Bhave S, Rodriguez V, Poterucha T, Mutasa S, et al. Eur Heart J. 2024. DOI for “Deep learning to detect left ventricular structural abnormalities in chest X-rays”
Deep learning-assisted diffusion tensor imaging for evaluation of the physis and metaphysis.Duong PT, Santos L, Hsu HY, Jambawalikar S, Mutasa S, Nguyen MK, Guariento A, Jaramillo D. J Imaging Inform Med. 2024. DOI for “Deep learning-assisted diffusion tensor imaging for evaluation of the physis and metaphysis”
Deep learning-assisted identification of femoroacetabular impingement (FAI) on routine pelvic radiographs.Hoy MK, Desai V, Mutasa S, Hoy RC, Gorniak R, Belair JA. J Imaging Inform Med. 2024. DOI for “Deep learning-assisted identification of femoroacetabular impingement (FAI) on routine pelvic radiographs”

2023

Breast cancer risk prediction combining a convolutional neural network-based mammographic evaluation with clinical factors.Michel A, Ro V, McGuinness JE, Mutasa S, et al. Breast Cancer Res Treat. 2023. DOI for “Breast cancer risk prediction combining a convolutional neural network-based mammographic evaluation with clinical factors”

2022

Use of a convolutional neural network-based mammographic evaluation to predict breast cancer recurrence among women with hormone receptor-positive operable breast cancer.McGuinness JE, Ro V, Mutasa S, et al. Breast Cancer Res Treat. 2022;194(1):35-47. DOI for “Use of a convolutional neural network-based mammographic evaluation to predict breast cancer recurrence among women with hormone receptor-positive operable breast cancer”
Deep learning prediction of axillary lymph node status using ultrasound images.Sun S, Mutasa S, Liu MZ, et al. Comput Biol Med. 2022;143:105250. DOI for “Deep learning prediction of axillary lymph node status using ultrasound images”

2021

Clinical artificial intelligence applications: musculoskeletal.Mutasa S, Yi PH. Radiol Clin North Am. 2021;59(6):1013-1026. DOI for “Clinical artificial intelligence applications: musculoskeletal”
AI MSK clinical applications: orthopedic implants.Yi PH, Mutasa S, Fritz J. Skeletal Radiol. 2021. DOI for “AI MSK clinical applications: orthopedic implants”
Understanding artificial intelligence based radiology studies: CNN architecture.Mutasa S, Sun S, Ha R. Clin Imaging. 2021;80:72-76. DOI for “Understanding artificial intelligence based radiology studies: CNN architecture”
Deciphering musculoskeletal artificial intelligence for clinical applications: how do I get started?Mutasa S, Yi PH. Skeletal Radiol. 2021. DOI for “Deciphering musculoskeletal artificial intelligence for clinical applications: how do I get started?”
Deep learning of computed tomography virtual wedge resection for prediction of histologic usual interstitial pneumonitis.Shaish H, Ahmed FS, Lederer D, et al., Mutasa S. Ann Am Thorac Soc. 2021;18(1):51-59. DOI for “Deep learning of computed tomography virtual wedge resection for prediction of histologic usual interstitial pneumonitis”

2020

Dynamic changes of convolutional neural network-based mammographic breast cancer risk score among women undergoing chemoprevention treatment.Manley H, Mutasa S, Chang P, et al. Clin Breast Cancer. 2020. DOI for “Dynamic changes of convolutional neural network-based mammographic breast cancer risk score among women undergoing chemoprevention treatment”
Development and external validation of a prognostic tool for COVID-19 critical disease.Chow DS, Glavis-Bloom J, Soun JE, et al., Mutasa S, et al. PLoS One. 2020;15(12):e0242953. DOI for “Development and external validation of a prognostic tool for COVID-19 critical disease”
A novel CNN algorithm for pathological complete response prediction using an I-SPY TRIAL breast MRI database.Liu MZ, Mutasa S, Chang P, et al. Magn Reson Imaging. 2020;73:148-151. DOI for “A novel CNN algorithm for pathological complete response prediction using an I-SPY TRIAL breast MRI database”
Advanced deep learning techniques applied to automated femoral neck fracture detection and classification.Mutasa S, Varada S, Goel A, Wong TT, Rasiej MJ. J Digit Imaging. 2020;33(5):1209-1217. DOI for “Advanced deep learning techniques applied to automated femoral neck fracture detection and classification”
Prospective analysis using a novel CNN algorithm to distinguish atypical ductal hyperplasia from ductal carcinoma in situ in breast.Mutasa S, Chang P, Nemer J, et al. Clin Breast Cancer. 2020;20(6):e757-e760. DOI for “Prospective analysis using a novel CNN algorithm to distinguish atypical ductal hyperplasia from ductal carcinoma in situ in breast”
Rethinking Greulich and Pyle: a deep learning approach to pediatric bone age assessment using pediatric trauma hand radiographs.Pan I, Baird GL, Mutasa S, et al. Radiol Artif Intell. 2020;2(4):e190198. DOI for “Rethinking Greulich and Pyle: a deep learning approach to pediatric bone age assessment using pediatric trauma hand radiographs”
Channel width optimized neural networks for liver and vessel segmentation in liver iron quantification.Liu M, Vanguri R, Mutasa S, et al. Comput Biol Med. 2020;122:103798. DOI for “Channel width optimized neural networks for liver and vessel segmentation in liver iron quantification”
Understanding artificial intelligence based radiology studies: what is overfitting?Mutasa S, Sun S, Ha R. Clin Imaging. 2020;65:96-99. DOI for “Understanding artificial intelligence based radiology studies: what is overfitting?”
Fully automated segmentation algorithm for perihematomal edema volumetry after spontaneous intracerebral hemorrhage.Ironside N, Chen CJ, Mutasa S, et al. Stroke. 2020;51(3):815-823. DOI for “Fully automated segmentation algorithm for perihematomal edema volumetry after spontaneous intracerebral hemorrhage”
Potential role of convolutional neural network based algorithm in patient selection for DCIS observation trials using a mammogram dataset.Mutasa S, Chang P, Van Sant EP, et al. Acad Radiol. 2020;27(6):774-779. DOI for “Potential role of convolutional neural network based algorithm in patient selection for DCIS observation trials using a mammogram dataset”

2019

Fully automated segmentation algorithm for hematoma volumetric analysis in spontaneous intracerebral hemorrhage.Ironside N, Chen CJ, Mutasa S, et al. Stroke. 2019;50(12):3416-3423. DOI for “Fully automated segmentation algorithm for hematoma volumetric analysis in spontaneous intracerebral hemorrhage”
Global trend in artificial intelligence-based publications in radiology from 2000 to 2018.West E, Mutasa S, Zhu Z, Ha R. AJR Am J Roentgenol. 2019;213(6):1204-1206. DOI for “Global trend in artificial intelligence-based publications in radiology from 2000 to 2018”
Prior to initiation of chemotherapy, can we predict breast tumor response? Deep learning convolutional neural networks approach using a breast MRI tumor dataset.Ha R, Chin C, Karcich J, et al., Mutasa S, et al. J Digit Imaging. 2019;32(5):693-701. DOI for “Prior to initiation of chemotherapy, can we predict breast tumor response? Deep learning convolutional neural networks approach using a breast MRI tumor dataset”
Eye tracking for deep learning segmentation using convolutional neural networks.Stember JN, Celik H, Krupinski E, Chang PD, Mutasa S, et al. J Digit Imaging. 2019;32(4):597-604. DOI for “Eye tracking for deep learning segmentation using convolutional neural networks”
Accuracy of distinguishing atypical ductal hyperplasia from ductal carcinoma in situ with convolutional neural network-based machine learning approach using mammographic image data.Ha R, Mutasa S, Sant EPV, et al. AJR Am J Roentgenol. 2019. DOI for “Accuracy of distinguishing atypical ductal hyperplasia from ductal carcinoma in situ with convolutional neural network-based machine learning approach using mammographic image data”
Predicting breast cancer molecular subtype with MRI dataset utilizing convolutional neural network algorithm.Ha R, Mutasa S, Karcich J, et al. J Digit Imaging. 2019;32(2):276-282. DOI for “Predicting breast cancer molecular subtype with MRI dataset utilizing convolutional neural network algorithm”
Prediction of lymph node maximum standardized uptake value in patients with cancer using a 3D convolutional neural network: a proof-of-concept study.Shaish H, Mutasa S, Makkar J, et al. AJR Am J Roentgenol. 2019;212(2):238-244. DOI for “Prediction of lymph node maximum standardized uptake value in patients with cancer using a 3D convolutional neural network: a proof-of-concept study”
Convolutional neural network using a breast MRI tumor dataset can predict Oncotype Dx recurrence score.Ha R, Chang P, Mutasa S, et al. J Magn Reson Imaging. 2019;49(2):518-524. DOI for “Convolutional neural network using a breast MRI tumor dataset can predict Oncotype Dx recurrence score”
Fully automated convolutional neural network method for quantification of breast MRI fibroglandular tissue and background parenchymal enhancement.Ha R, Chang P, Mema E, Mutasa S, et al. J Digit Imaging. 2019;32(1):141-147. DOI for “Fully automated convolutional neural network method for quantification of breast MRI fibroglandular tissue and background parenchymal enhancement”
Convolutional neural network based breast cancer risk stratification using a mammographic dataset.Ha R, Chang P, Karcich J, Mutasa S, et al. Acad Radiol. 2019;26(4):544-549. DOI for “Convolutional neural network based breast cancer risk stratification using a mammographic dataset”

2018

Predicting post neoadjuvant axillary response using a novel convolutional neural network algorithm.Ha R, Chang P, Karcich J, Mutasa S, et al. Ann Surg Oncol. 2018;25(10):3037-3043. DOI for “Predicting post neoadjuvant axillary response using a novel convolutional neural network algorithm”
Axillary lymph node evaluation utilizing convolutional neural networks using MRI dataset.Ha R, Chang P, Karcich J, Mutasa S, et al. J Digit Imaging. 2018;31(6):851-856. DOI for “Axillary lymph node evaluation utilizing convolutional neural networks using MRI dataset”
MABAL: a novel deep-learning architecture for machine-assisted bone age labeling.Mutasa S, Chang PD, Ruzal-Shapiro C, Ayyala R. J Digit Imaging. 2018;31(4):513-519. DOI for “MABAL: a novel deep-learning architecture for machine-assisted bone age labeling”

2017 & earlier

Images in COPD: bullous emphysema with mycetoma.Mutasa S. Chronic Obstr Pulm Dis. 2017;4(2):168-171. DOI for “Images in COPD: bullous emphysema with mycetoma”
Genetic dysregulation in recurrent respiratory papillomatosis.Rodman R, Mutasa S, Dupuis C, Spratt H, Underbrink M. Laryngoscope. 2014;124(8):E320-5. DOI for “Genetic dysregulation in recurrent respiratory papillomatosis”

Podium

Oral presentations

30 presentations, including two invited RSNA courses.

INVITED

Interactive Artificial Intelligence for Non-Coders: Object Localization and Image Segmentation.Invited faculty, educational course. RSNA Annual Meeting (virtual), 2020.
Hands-on Artificial Intelligence for Non-Coders: Object Localization and Image Segmentation.Invited faculty, educational course. RSNA Annual Meeting, Chicago, 2019.

2022

Can we trust saliency maps to “explain” deep learning algorithms for musculoskeletal radiograph abnormality detection?Venkatesh K, Mutasa S, Moore F, Sulam J, Yi PH. SIIM Annual Meeting, 2022.

2021

Determining the signal characteristics of disk herniations that are more likely to resorb.Sax AJ, Xia E, Alfaqih M, Mutasa S, Zoga AC, Deely D, Belair J, Kamel S, Morrison WB. Society of Skeletal Radiology, 2021.
Fracture incidence before, during, and after COVID-19 lockdown.Sax AJ, Xia E, Deely D, Karasick D, Alfaqih M, Mutasa S, Morrison WB. Society of Skeletal Radiology, 2021.

2020

Can AI improve the detection of scaphoid fractures that have equivocal findings on radiograph?Mutasa S, Igbinoba Z, Ubanwa A, Lala S, Wong T. RSNA, 2020.
LVHNet: detecting cardiac structural abnormalities with chest X-rays.Elias P, Shreyas B, Mutasa S, Rodriguez V, Leb J, Poterucha T, Perotte A. Medical Imaging Meets NeurIPS, 2020.
Virtual pulmonary wedge resection: a CNN-based tool for classifying UIP vs non-UIP ILD on imaging.Mutasa S, D’Souza B, Salvatore M, Lederer D, Shaish H. SIIM Annual Meeting, 2020.
Can a CNN-based algorithm aid in identification of patients with BRCA mutation using a mammographic dataset?Mutasa S, Chang P, Ha R. SIIM Annual Meeting, 2020.
Physeal diffusion tensor tractography ROI automation with a 3D convolutional neural network.Mutasa S, Liu M, Jambawalikar S, Jaramillo D. Society for Pediatric Radiology, 2020.

2019

A prospective analysis distinguishing ADH from DCIS using a CNN.Mutasa S, Chang P, Nemer J, Liu M, Jambawalikar S, Ha R. RSNA, 2019.
The effect of chemopreventive agents on a CNN-based breast cancer risk model.Mutasa S, Manley H, Pascual Van Sant E, Chang P, Crew K, Ha R. RSNA, 2019.
Rethinking Greulich & Pyle: deep learning models for pediatric bone age assessment.Pan I, Mutasa S, Merck D, Ruzal-Shapiro C, Ayyala R, Swenson D. RSNA, 2019.
Effect of chemoprevention on a CNN-based breast cancer risk model using a mammographic dataset of women with ADH, DCIS, or LCIS.Manley H, Mutasa S, Ha R, Crew K. San Antonio Breast Cancer Symposium, 2019.
Synchronous distributed deep learning for medical imaging.Mutasa S, Chang P. Conference on Machine Intelligence in Medical Imaging (C-MIMI), 2019.
Predicting post-neoadjuvant axillary response using a CNN.Mutasa S, Karcich J, Chang P, Pascual Van Sant E, Connolly E, Chin C, Liu M, Taback B, Ha R. American Roentgen Ray Society, 2019.
Can deep learning differentiate HCC from non-HCC on MRI?Mutasa S, Kierans A, Lu L, Hecht A. Society of Abdominal Radiology, 2019.
A fully automated segmentation algorithm for the volumetric analysis of perihematomal edema in patients with spontaneous intracerebral hemorrhage.Ironside N, Mutasa S, Jambawalikar S, Vanguri R, Lignelli A, Connolly E. International Stroke Conference, 2019.

2018

Potential role of CNN-based algorithms in patient selection for DCIS observation trials.Mutasa S, Chang P, Karcich J, Pascual Van Sant E, Sun M, Liu M, Ha R. RSNA, 2018.
Utilizing deep learning to obtain preoperative measurements from radiographs in patients with femoroacetabular impingement.Mutasa S, Igbinoba Z, Rasiej M, Wong T. RSNA, 2018.
Convolutional neural network-based breast cancer risk stratification using a mammographic dataset.Mutasa S, Chang P, Karcich J, Pascual Van Sant E, Sun M, Liu M, Ha R. RSNA, 2018.
Advanced data augmentation techniques for medical deep learning applications.Mutasa S, Goel A, Varada S, Liu A, Wong T, Rasiej M. Conference on Machine Intelligence in Medical Imaging (C-MIMI), 2018.
Automated femoral neck fracture detection and classification utilizing advanced deep learning techniques.Mutasa S, Varada S, Goel A, Wong T, Rasiej M. International Skeletal Society, 2018.
Predicting glioblastoma tumor recurrence vs pseudoprogression utilizing convolutional neural networks.Mutasa S, Genestine W, Zanazzi G, Delgado N, Iwamoto F, Chang P, Canoll P, Lassman A, Bruce J, Lignelli A. American Society of Neuroradiology, 2018.
Predicting breast cancer molecular subtype with an MRI data set.Mutasa S, Jambawalikar S, Ha R. SIIM Annual Meeting, 2018.
A deep learning approach to automated cerebral aneurysm detection on magnetic resonance angiograms with application to automatically predict the size of basilar tip aneurysms.Stember JN, Chang PD, Stember DM, Liu M, Marfatia S, Vanguri R, Mutasa S, Filippi R, Meyers P, Lignelli A, Jambawalikar S. American Society of Neuroradiology, 2018.
Augmented reality guidance for cerebral angiography.Loeb G, Sadri S, Grinshpoon A, Carroll J, Cooper C, Elvezio C, Mutasa S, Mandigo G, Lavine S, Weintraub J, Einstein A, Feiner S, Meyers P. Society of Interventional Radiology, 2018.
Motion tolerant super resolution reconstruction from multi-stack MRI data.Jambawalikar S, Vanguri R, Mutasa S, Liu M, Shaish H. International Society for Magnetic Resonance in Medicine, 2018.

2017 & earlier

Deep learning through convolutional neural networks using a breast MRI tumor dataset can predict neoadjuvant chemotherapy response.Karcich J, Jambawalikar S, Chang PD, Chin C, Jadeja P, Liu M, Wynn R, Mutasa S, Pascual Van Sant E, Connolly E, Ha R. RSNA, 2017.
Gene expression profiles in recurrent respiratory papillomatosis.Rodman R, Mutasa S, Dupuis C, Spratt H, Underbrink M. Triological Society, COSM, 2013.

Exhibits

Poster presentations

9 posters.

Physeal stress injuries vs Salter-Harris fractures: the role of the perichondrium.Mutasa S, Huang S, Taol L, Jaramillo D. RSNA, 2020.
Key concepts in imaging informatics: a curriculum for radiologists.Liu A, Goel A, Marfatia S, Mutasa S. Association of University Radiologists, 2019.
A fully automated segmentation algorithm for the volumetric analysis of perihematomal edema in patients with spontaneous intracerebral hemorrhage.Ironside N, Mutasa S, Roh D, Mayer S, Chen C, Jambawalikar S, Vanguri R, Lignelli A, Connolly ES. International Stroke Conference, 2019.
Automated femoral neck fracture detection and classification utilizing advanced deep learning techniques.Mutasa S, Varada S, Goel A, Wong T, Rasiej M. RSNA, 2018.
Tracking patient movement for augmented reality assisted fluoroscopic procedures using convolutional neural networks.Mutasa S, Wu B, Loeb G, Chang P, Marfatia S, Meyers P. SIIM Annual Meeting, 2018.
Distinguishing atypical ductal hyperplasia from ductal carcinoma in situ: convolutional neural network-based machine learning approach using mammographic image data.Mutasa S, Jambawalikar S, Lu M, Pascual Van Sant E, Ha R. American Society of Breast Surgeons, 2018.
Patient selection for DCIS observation trials (COMET and LORIS): are they truly low risk?Patel G, Pascual Van Sant E, Mutasa S, Wynn R, Ha R. American Roentgen Ray Society, 2018.
Categorizing peripheral neuropathy utilizing deep learning.Okonkwo N, Mutasa S, Ho C. United States & Canadian Academy of Pathology, 2018.
Targeting molecular mechanisms for the treatment of recurrent respiratory papillomatosis.Mutasa S, Underbrink M. UTMB Medical Student Summer Research Fair, 2012.

Service & recognition

Leadership & honors

LEADERSHIP & SERVICE

Machine Learning Research Tools SubcommitteeSociety for Imaging Informatics in Medicine (SIIM), 2019 – present
Clinical Lead, Arterys AI platform integrationNewYork-Presbyterian Hospital, 2018 – 2020
Chief Resident, Diagnostic RadiologyColumbia University Irving Medical Center, 2017 – 2019
Associate Editor for RadiologyMedscape & WebMD, Medscape Consult
Manuscript reviewerJ Digit Imaging (2018–), Scientific Reports (2019–), Radiotherapy & Oncology (2020–)
Scientific session moderator“Machine Learning: MSK & Women’s Imaging,” SIIM Annual Meeting, 2018
NYP COVID-19 responseClinical resident in a repurposed OR ICU and portable X-ray technologist, 2020
Clinical volunteer, St. Vincent’s Free ClinicHealthcare for the indigent population of Galveston, TX, 2012 – 2015
MembershipsRSNA, ACR, ARRS, SSR, AMA, Alpha Omega Alpha

GRANTS & HONORS

NIH StARR R38 Award, Columbia CAPRI ProgramOne funded year of resident research on AI-based breast cancer risk models, $60k, 2019 – 2020. news
Columbia Herbert Irving Cancer Center Velocity GrantCo-investigator, translational research, $100k, 2019 – 2020
Roger A. Bauman, MD AwardBest trainee paper in imaging informatics, SIIM, 2018
RSNA Travel Award2018
American Roentgen Ray Society Scholar Grant2017
Thayer Award for Excellence in Teaching2016
Alpha Omega Alpha Honor Medical SocietyUTMB, 2015
NIH NHLBI Medical Student Research Grant2012
Mayo Clinic School of Graduate Education Scholarship2014