Adrian Vasile Dalca

CV | scholar | linkedin | twitter | github

Assistant Professor, A.A. Martinos Center for Biomedical Imaging
Massachusetts General Hospital, Harvard Medical School

Research Scientist, Computer Science and Artificial Intelligence Lab
EECS, Massachusetts Institute of Technology

32 Vassar St, 32-G904, Cambridge, MA, 02139 
adalca at mit dot edu

My research focuses on machine learning techniques and probabilistic models with a frequent focus on medical image analysis, computer vision, and healthcare. Our group is excited about model that enable new tasks and new possibilities, especially in healthcare.

I was a postdoctoral fellow at CSAIL, MIT and MGH, Harvard Medical School, working with Mert Sabuncu and John Guttag. I completed my PhD in the Medical Vision Group, CSAIL, EECS, MIT, advised by Polina Golland.

My wife, Monica, completed her PhD at MIT in the Biology department doing exciting research in cancer biology.

Recent Updates

Oct 2023 Adrian received the MICCAI Young Scientist Impact Award for substantial publication impact in the field in the last 5 years.
Sep 2023 Our lab received an NIH R01 Grant for building robust/invariant DL methods for image analysis!
Sep 2023 Jose J.G. Ortiz will present Scale-space Hypernetworks at NeurIPS 2023!
July 2023 Check out UniverSeg, a universal medical image segmentation tool that does not require retraining or interaction for each new segmentation task! Will be presented at ICCV 2023 by co-first-authors Victor Butoi and Jose Javier Ortiz.
Feb 2023 Neuralizer, a tool for Neuroimage Analysis that does not require re-training for every neuroimaging task, has been accepted at CVPR!
Apr 2022 HyperMorph -- our image registration method for avoiding hyperparameter tuning has been published in MELBA!
Mar 2022 SynthMorph -- a general machine learning tool for image registration -- has been published in TMI!
Mar 2022 topoFit -- volumetric and surface neural networks for surface fitting -- will be presented by Andrew Hoopes at MIDL2022!
Feb 2022 Our work was featured on Computer Vision News!
Sep 2021 Malte Hoffmann's work on SynthMorph: modality-agnostic neural networks for registration now accepted at TMI!
July 2021 Congrats to Andrew for his runner-up for the Erbsmann prize (best paper/presentation) for his HyperMorph presentation at IPMI 2021!
Feb 2021 Check out our new method on avoiding hyperparameter tuning, applied first on registration in HyperMorph: learning registration-hyperparameter-free hypernetworks for registration ! Led by Andrew Hoopes, accepted at IPMI 2021!
Feb 2021 Check out Aniruddh Raghu's work on explaining healthcare models with medical evidence, to be presented at ACM CHIL 2021!
Jan 2021 Malte Hoffmann's work on SynthMorph: modality-agnostic neural networks for registration will be presented at ISBI 2021!
Sep 2020 Come to the learn2reg Challenge at MICCAI 2020.
June 2020 Check our work on optimizing the under-sampling pattern in MRI, accepted at IEEE TCI!
May 2020 Check our work on expanding SynthSeg to partial voluming, to be presented at MICCAI 2020! Led by Benjamin Billot, the method can segment very thick and sparse volumes (7mm MR scans)!
May 2020 Check our the NeurIPS Machine Learning for Health 2019 Workshop proceedings.
Apr 2020 Check our work on a learning strategy for contrast-agnostic segmentation, to be presented at MIDL 2020! Led by Benjamin Billot, the method can segment any brain imaging modality, even those never seen before.
Apr 2020 Marianne Rakic's short paper, to be presented at MIDL 2020, describes an antomical prediction model using subject-specific medical information (clinical status, genetics, etc)!
Feb 2020 Amy Zhao shows how to synthesize painting timelapses at CVPR 2020.
Feb 2020 Check out Katie Lewis's work on porting VoxelMorph to a clinical images at ACM CHIL 2020.
Feb 2020 Check out He Sun's work on learning a sensing strategy for computational imaging, accepted at ICCP 2020, with Katie Bouman.
Oct 2019 Come to the learn2reg Tutorial on Thursday to learn about deep learning in registration at MICCAI 2019.
Oct 2019 Check out our analysis, led by Brett Beaulieu-Jones, on the trends in Machine Learning for Health research at JAMA Network Open!
Sep 2019 Check out our paper on a learning method for building deformable templates jointly with learning registration network, accepted at NeurIPS 2019!
July 2019 Check out the new task of Visual Deprojection, led by Guha Balakrishnan, in which we describe a problem and solution for Recovery of Collapsed Dimensions in Images and Videos. Accepted at ICCV 2019.
July 2019 Our analysis of the diffeomorphic Voxelmorph variant, along with its extension to surface registration, is accepted at MedIA: Medical Image Analysis.
June 2019 Our recent work on combining deep learning with classical baysian models for unsupervised segmentation has been accepted to MICCAI 2019.
May 2019 Markus Schirmer's long-time effort on Clinical WMH Quantification accepted at NeuroImage: Clinical.
Mar 2019 Check out a new strategy to learn data augmentation for one-shot medical image segmentation, led by Amy Zhao. Accepted as Oral Presentation at CVPR 2019.
Feb 2019 Check out a joint learning framework for acquisition and reconstruction, led by Cagla Bahadir. Accepted at IPMI 2019.
Dec 2018 Our analysis of the Voxelmorph framework is accepted at IEEE TMI: Transactions on Medical Imaging.
Nov 2018 Several papers accepted at NeurIPS Workshops: ML4H: Machine Learning for Health and MED-NeurIPS: Medical Imaging Meets NeurIPS.
Sep 2018 Voxelmorph (Probabilistic Diffeomorphic registration) is a Finalist for best paper (Young Scientist) award at MICCAI!
Sep 2018 Our paper, led by Paolo Casale, on using Gaussian Process Priors in Variational Autoencoders, accepted at NeurIPS!
Jul 2018 Our paper on Medical Image Imputation from Image Collections accepted at IEEE TMI: Transactions on Medical Imaging.
Jul 2018 Our paper led by Danielle Pace on Iterative Segmentation from Limited Training Data accepted at MICCAI-DLMIA: Deep Learning in Medical Image Analysis!
Jun 2018 News articles about VoxelMorph and pose warping [1] [2] came out during the week of cvpr where we presented three papers.
Apr 2018 Our paper "Unsupervised Learning for Fast Probabilistic Diffeomorphic Registration" is an early-accept at MICCAI 2018, where we will present it as an oral presentation! We develop a connection between classical probabilistic diffeomorphic models and learning to register with Convolutional Neural Networks.
Apr 2018 Our paper led by Katie Bouman on Reconstructing Video of Time-varying Celestial Objects accepted at IEEE TCI: Transactions on Computational Imaging!
Mar 2018 We will be organizing a new technical workshop at MICCAI this year focusing on integrating medical imaging and non-imaging modalities to answer novel clinical and healthcare challenges.
Feb 2018 Three papers accepted at CVPR 2018!

Webdesign: Adrian Dalca. Based on: MiniFolio