Data-driven MedicineConfidence-based anomaly detection for automatic quantification of impairment and disease severityWe present a method to perform automatic assessment of impairment and disease severity using AI models trained only on healthy individuals. The COnfidence-Based chaRacterization of Anomalies (COBRA) score exploits the decrease in confidence of these models when processing data from impaired or diseased patients to quantify their deviation from the healthy population. We show that the COBRA score is strongly correlated with expert-based metrics for two different medical conditions (stroke and knee osteoarthritis) and three data modalities (wearable sensors, video and magnetic-resonance scans).
Early detection of Alzheimer's via deep learningWe design 3D convolutional neural network to detect Alzheimer's Disease using structural brain MRI scans. This research is supported by Alzheimer’s Association grant AARG-NTF-21-848627.
Weakly-supervised segmentation for breast-cancer diagnosisWeakly-supervised segmentation is a framework to interpret the output of deep-learning models. This approach produces saliency maps highlighting the regions of the input most relevant to the classification task (e.g. malignant lesions in mammograms) using only image-level labels (e.g. whether the patient has cancer or not) during training. When applied to high-resolution images, existing methods produce low-resolution saliency maps. This is problematic in applications such as breast cancer where lesions can be very small in relation to the image size. To address this, we introduce a novel neural network architecture for weakly-supervised segmentation of high-resolution images. The proposed model selects regions of interest via coarse-level localization, and then performs fine-grained segmentation of those regions. We apply this model to breast cancer diagnosis with screening mammography, and validate it on a large clinically-realistic dataset.
Automatic prognostics of COVID-19During the COVID-19 pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a system for automatic prediction of deterioration risk with two components: a deep neural network that processes chest X-ray images, and a gradient boosting model applied to routine clinical variables. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at NYU Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time.
Magnetic resonance fingerprintingMagnetic resonance fingerprinting (MRF) is a recently-developed technique for quantitative estimation of tissue parameters in the human body. These parameters have great potential as biomarkers for various pathologies, and allow to synthesize images with standardized contrasts. Our work focuses on optimizing measurement design, performing recovery using deep learning, and adapting the MRF framework to account for the presence of several tissues in each voxel. This research was supported by NIH grant R21 EB027241.
Analysis of infant-sleep patternsWe propose a nonparametric model for time series with missing data based on regularized nonnegative low-rank matrix factorization. The model expresses each instance in a set of time series as a linear combination of a small number of shared basis functions. The methodology is applied to a large real-world dataset of infant-sleep data gathered by caregivers with a mobile-phone app.
Parallel magnetic-resonance imaging and compressed sensingUndersampling images in the frequency domain enables accelerated acquisition in magnetic resonance imaging. Here we study how to combine two complementary approaches: parallel imaging (i.e. using multiple coils with different sensitivities to gather the data), and compressed sensing (i.e. randomizing the sampling pattern, and exploiting image sparsity in a transform domain).
|