Narges Razavian

Kangning Liu, Weicheng Zhu, Yiqiu Shen et al.

Learning representations for individual instances when only bag-level labels are available is a fundamental challenge in multiple instance learning (MIL). Recent works have shown promising results using contrastive self-supervised learning (CSSL), which learns to push apart representations corresponding to two different randomly-selected instances. Unfortunately, in real-world applications such as medical image classification, there is often class imbalance, so randomly-selected instances mostly belong to the same majority class, which precludes CSSL from learning inter-class differences. To address this issue, we propose a novel framework, Iterative Self-paced Supervised Contrastive Learning for MIL Representations (ItS2CLR), which improves the learned representation by exploiting instance-level pseudo labels derived from the bag-level labels. The framework employs a novel self-paced sampling strategy to ensure the accuracy of pseudo labels. We evaluate ItS2CLR on three medical datasets, showing that it improves the quality of instance-level pseudo labels and representations, and outperforms existing MIL methods in terms of both bag and instance level accuracy. Code is available at https://github.com/Kangningthu/ItS2CLR

Weicheng Zhu, Carlos Fernandez-Granda, Narges Razavian

Lung squamous cell carcinoma (LSCC) has a high recurrence and metastasis rate. Factors influencing recurrence and metastasis are currently unknown and there are no distinct histopathological or morphological features indicating the risks of recurrence and metastasis in LSCC. Our study focuses on the recurrence prediction of LSCC based on H&E-stained histopathological whole-slide images (WSI). Due to the small size of LSCC cohorts in terms of patients with available recurrence information, standard end-to-end learning with various convolutional neural networks for this task tends to overfit. Also, the predictions made by these models are hard to interpret. Histopathology WSIs are typically very large and are therefore processed as a set of smaller tiles. In this work, we propose a novel conditional self-supervised learning (SSL) method to learn representations of WSI at the tile level first, and leverage clustering algorithms to identify the tiles with similar histopathological representations. The resulting representations and clusters from self-supervision are used as features of a survival model for recurrence prediction at the patient level. Using two publicly available datasets from TCGA and CPTAC, we show that our LSCC recurrence prediction survival model outperforms both LSCC pathological stage-based approach and machine learning baselines such as multiple instance learning. The proposed method also enables us to explain the recurrence histopathological risk factors via the derived clusters. This can help pathologists derive new hypotheses regarding morphological features associated with LSCC recurrence.

Weicheng Zhu, Sheng Liu, Carlos Fernandez-Granda et al.

Self-supervised learning (SSL) has emerged as a powerful technique for learning rich representations from unlabeled data. The data representations are able to capture many underlying attributes of data, and be useful in downstream prediction tasks. In real-world settings, spurious correlations between some attributes (e.g. race, gender and age) and labels for downstream tasks often exist, e.g. cancer is usually more prevalent among elderly patients. In this paper, we investigate SSL in the presence of spurious correlations and show that the SSL training loss can be minimized by capturing only a subset of the conspicuous features relevant to those sensitive attributes, despite the presence of other important predictive features for the downstream tasks. To address this issue, we investigate the learning dynamics of SSL and observe that the learning is slower for samples that conflict with such correlations (e.g. elder patients without cancer). Motivated by these findings, we propose a learning-speed aware SSL (LA-SSL) approach, in which we sample each training data with a probability that is inversely related to its learning speed. We evaluate LA-SSL on three datasets that exhibit spurious correlations between different attributes, demonstrating that it improves the robustness of pretrained representations on downstream classification tasks.

Haresh Rengaraj Rajamohan, Xiang Gao, Weicheng Zhu et al.

While large-scale pretraining has revolutionized language modeling, its potential remains underexplored in healthcare with structured electronic health records (EHRs). We present RAVEN, a novel generative pretraining strategy for sequential EHR data based on Recurrence-Aware next-Visit EveNt prediction. Leveraging a dataset of over one million unique individuals, our model learns to autoregressively generate tokenized clinical events for the next visit conditioned on patient history. We introduce regularization on predicting repeated events and highlight a key pitfall in EHR-based foundation model evaluations: repeated event tokens can inflate performance metrics when new onsets are not distinguished from subsequent occurrences. Furthermore, we empirically investigate the scaling behaviors in a data-constrained, compute-saturated regime, showing that simply increasing model size is suboptimal without commensurate increases in data volume. We evaluate our model via zero-shot prediction for forecasting the incidence of a diverse set of diseases, where it rivals fully fine-tuned representation-based Transformer models and outperforms widely used simulation-based next-token approaches. Finally, without additional parameter updates, we show that RAVEN can generalize to an external patient cohort under lossy clinical code mappings and feature coverage gaps.

Long Chen, Liben Chen, Binfeng Xu et al.

The aging population of the U.S. drives the prevalence of Alzheimer's disease. Brookmeyer et al. forecasts approximately 15 million Americans will have either clinical AD or mild cognitive impairment by 2060. In response to this urgent call, methods for early detection of Alzheimer's disease have been developed for prevention and pre-treatment. Notably, literature on the application of deep learning in the automatic detection of the disease has been proliferating. This study builds upon previous literature and maintains a focus on leveraging multi-modal information to enhance automatic detection. We aim to predict the stage of the disease - Cognitively Normal (CN), Mildly Cognitive Impairment (MCI), and Alzheimer's Disease (AD), based on two different types of brain MRI scans. We design an AlexNet-based deep learning model that learns the synergy of complementary information from both T1 and FLAIR MRI scans.

Sheng Liu, Long Chen, Zeyun Zhao et al.

Modern clinical practice increasingly depends on reasoning over heterogeneous, evolving, and incomplete patient data. Although recent advances in multimodal foundation models have improved performance on various clinical tasks, most existing models remain static, opaque, and poorly aligned with real-world clinical workflows. We present Cerebra, an interactive multi-agent AI team that coordinates specialized agents for EHR, clinical notes, and medical imaging analysis. These outputs are synthesized into a clinician-facing dashboard that combines visual analytics with a conversational interface, enabling clinicians to interrogate predictions and contextualize risk at the point of care. Cerebra supports privacy-preserving deployment by operating on structured representations and remains robust when modalities are incomplete. We evaluated Cerebra using a massive multi-institutional dataset spanning 3 million patients from four independent healthcare systems. Cerebra consistently outperformed both state-of-the-art single-modality models and large multimodal language model baselines. In dementia risk prediction, it achieved AUROCs up to 0.80, compared with 0.74 for the strongest single-modality model and 0.68 for language model baselines. For dementia diagnosis, it achieved an AUROC of 0.86, and for survival prediction, a C-index of 0.81. In a reader study with experienced physicians, Cerebra significantly improved expert performance, increasing accuracy by 17.5 percentage points in prospective dementia risk estimation. These results demonstrate Cerebra's potential for interpretable, robust decision support in clinical care.

Aakash Kaku, Chaitra V. Hegde, Jeffrey Huang et al.

Quantitative, volumetric analysis of Magnetic Resonance Imaging (MRI) is a fundamental way researchers study the brain in a host of neurological conditions including normal maturation and aging. Despite the availability of open-source brain segmentation software, widespread clinical adoption of volumetric analysis has been hindered due to processing times and reliance on manual corrections. Here, we extend the use of deep learning models from proof-of-concept, as previously reported, to present a comprehensive segmentation of cortical and deep gray matter brain structures matching the standard regions of aseg+aparc included in the commonly used open-source tool, Freesurfer. The work presented here provides a real-life, rapid deep learning-based brain segmentation tool to enable clinical translation as well as research application of quantitative brain segmentation. The advantages of the presented tool include short (~1 minute) processing time and improved segmentation quality. This is the first study to perform quick and accurate segmentation of 102 brain regions based on the surface-based protocol (DMK protocol), widely used by experts in the field. This is also the first work to include an expert reader study to assess the quality of the segmentation obtained using a deep-learning-based model. We show the superior performance of our deep-learning-based models over the traditional segmentation tool, Freesurfer. We refer to the proposed deep learning-based tool as DARTS (DenseUnet-based Automatic Rapid Tool for brain Segmentation). Our tool and trained models are available at https://github.com/NYUMedML/DARTS

Weicheng Zhu, Haoxu Huang, Huanze Tang et al.

Head computed tomography (CT) imaging is a widely-used imaging modality with multitudes of medical indications, particularly in assessing pathology of the brain, skull, and cerebrovascular system. It is commonly the first-line imaging in neurologic emergencies given its rapidity of image acquisition, safety, cost, and ubiquity. Deep learning models may facilitate detection of a wide range of diseases. However, the scarcity of high-quality labels and annotations, particularly among less common conditions, significantly hinders the development of powerful models. To address this challenge, we introduce FM-CT: a Foundation Model for Head CT for generalizable disease detection, trained using self-supervised learning. Our approach pre-trains a deep learning model on a large, diverse dataset of 361,663 non-contrast 3D head CT scans without the need for manual annotations, enabling the model to learn robust, generalizable features. To investigate the potential of self-supervised learning in head CT, we employed both discrimination with self-distillation and masked image modeling, and we construct our model in 3D rather than at the slice level (2D) to exploit the structure of head CT scans more comprehensively and efficiently. The model's downstream classification performance is evaluated using internal and three external datasets, encompassing both in-distribution (ID) and out-of-distribution (OOD) data. Our results demonstrate that the self-supervised foundation model significantly improves performance on downstream diagnostic tasks compared to models trained from scratch and previous 3D CT foundation models on scarce annotated datasets. This work highlights the effectiveness of self-supervised learning in medical imaging and sets a new benchmark for head CT image analysis in 3D, enabling broader use of artificial intelligence for head CT-based diagnosis.

Haresh Rengaraj Rajamohan, Xiang Gao, Weicheng Zhu et al.

Large-scale pretraining has transformed modeling of language and other data types, but its potential remains underexplored in healthcare with structured electronic health records (EHRs). We present a novel generative pretraining strategy for sequential EHR data using next-visit event prediction. Our model learns to autoregressively generate various tokenized clinical events for the next visit based on patient history and inherently handles the joint prediction of heterogeneous data types. Additionally, we introduce regularization on predicting repeated events and highlight a key pitfall in EHR-based foundation model evaluations: repeated event tokens can inflate performance metrics when new onsets are not distinguished from subsequent occurrences. Our model is evaluated via zero-shot prediction for forecasting dementia and knee osteoarthritis incidence within 2 and 5 years, and the model performance rivals a fully fine-tuned masked pretrained Transformer baseline, demonstrating that our approach captures complex clinical dependencies without requiring costly task-specific fine-tuning.

Sheng Liu, Aakash Kaku, Weicheng Zhu et al.

Reliable probability estimation is of crucial importance in many real-world applications where there is inherent (aleatoric) uncertainty. Probability-estimation models are trained on observed outcomes (e.g. whether it has rained or not, or whether a patient has died or not), because the ground-truth probabilities of the events of interest are typically unknown. The problem is therefore analogous to binary classification, with the difference that the objective is to estimate probabilities rather than predicting the specific outcome. This work investigates probability estimation from high-dimensional data using deep neural networks. There exist several methods to improve the probabilities generated by these models but they mostly focus on model (epistemic) uncertainty. For problems with inherent uncertainty, it is challenging to evaluate performance without access to ground-truth probabilities. To address this, we build a synthetic dataset to study and compare different computable metrics. We evaluate existing methods on the synthetic data as well as on three real-world probability estimation tasks, all of which involve inherent uncertainty: precipitation forecasting from radar images, predicting cancer patient survival from histopathology images, and predicting car crashes from dashcam videos. We also give a theoretical analysis of a model for high-dimensional probability estimation which reproduces several of the phenomena evinced in our experiments. Finally, we propose a new method for probability estimation using neural networks, which modifies the training process to promote output probabilities that are consistent with empirical probabilities computed from the data. The method outperforms existing approaches on most metrics on the simulated as well as real-world data.

Daniel Jiwoong Im, Kyunghyun Cho, Narges Razavian

Domain adaptation and covariate shift are big issues in deep learning and they ultimately affect any causal inference algorithms that rely on deep neural networks. Causal effect variational autoencoder (CEVAE) is trained to predict the outcome given observational treatment data and it suffers from the distribution shift at test time. In this paper, we introduce uniform treatment variational autoencoders (UTVAE) that are trained with uniform treatment distribution using importance sampling and show that using uniform treatment over observational treatment distribution leads to better causal inference by mitigating the distribution shift that occurs from training to test time. We also explore the combination of uniform and observational treatment distributions with inference and generative network training objectives to find a better training procedure for inferring treatment effects. Experimentally, we find that the proposed UTVAE yields better absolute average treatment effect error and precision in the estimation of heterogeneous effect error than the CEVAE on synthetic and IHDP datasets.

Aakash Kaku, Sahana Upadhya, Narges Razavian

We show that bringing intermediate layers' representations of two augmented versions of an image closer together in self-supervised learning helps to improve the momentum contrastive (MoCo) method. To this end, in addition to the contrastive loss, we minimize the mean squared error between the intermediate layer representations or make their cross-correlation matrix closer to an identity matrix. Both loss objectives either outperform standard MoCo, or achieve similar performances on three diverse medical imaging datasets: NIH-Chest Xrays, Breast Cancer Histopathology, and Diabetic Retinopathy. The gains of the improved MoCo are especially large in a low-labeled data regime (e.g. 1% labeled data) with an average gain of 5% across three datasets. We analyze the models trained using our novel approach via feature similarity analysis and layer-wise probing. Our analysis reveals that models trained via our approach have higher feature reuse compared to a standard MoCo and learn informative features earlier in the network. Finally, by comparing the output probability distribution of models fine-tuned on small versus large labeled data, we conclude that our proposed method of pre-training leads to lower Kolmogorov-Smirnov distance, as compared to a standard MoCo. This provides additional evidence that our proposed method learns more informative features in the pre-training phase which could be leveraged in a low-labeled data regime.

Farah E. Shamout, Yiqiu Shen, Nan Wu et al.

During the coronavirus disease 2019 (COVID-19) pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3,661 patients, achieves an area under the receiver operating characteristic curve (AUC) of 0.786 (95% CI: 0.745-0.830) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at New York University Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.

Sheng Liu, Jonathan Niles-Weed, Narges Razavian et al.

We propose a novel framework to perform classification via deep learning in the presence of noisy annotations. When trained on noisy labels, deep neural networks have been observed to first fit the training data with clean labels during an "early learning" phase, before eventually memorizing the examples with false labels. We prove that early learning and memorization are fundamental phenomena in high-dimensional classification tasks, even in simple linear models, and give a theoretical explanation in this setting. Motivated by these findings, we develop a new technique for noisy classification tasks, which exploits the progress of the early learning phase. In contrast with existing approaches, which use the model output during early learning to detect the examples with clean labels, and either ignore or attempt to correct the false labels, we take a different route and instead capitalize on early learning via regularization. There are two key elements to our approach. First, we leverage semi-supervised learning techniques to produce target probabilities based on the model outputs. Second, we design a regularization term that steers the model towards these targets, implicitly preventing memorization of the false labels. The resulting framework is shown to provide robustness to noisy annotations on several standard benchmarks and real-world datasets, where it achieves results comparable to the state of the art.

Zachariah Zhang, Jingshu Liu, Narges Razavian

Clinical interactions are initially recorded and documented in free text medical notes. ICD coding is the task of classifying and coding all diagnoses, symptoms and procedures associated with a patient's visit. The process is often manual and extremely time-consuming and expensive for hospitals. In this paper, we propose a machine learning model, BERT-XML, for large scale automated ICD coding from EHR notes, utilizing recently developed unsupervised pretraining that have achieved state of the art performance on a variety of NLP tasks. We train a BERT model from scratch on EHR notes, learning with vocabulary better suited for EHR tasks and thus outperform off-the-shelf models. We adapt the BERT architecture for ICD coding with multi-label attention. While other works focus on small public medical datasets, we have produced the first large scale ICD-10 classification model using millions of EHR notes to predict thousands of unique ICD codes.

Weicheng Zhu, Narges Razavian

Electronic Health Records (EHR) are high-dimensional data with implicit connections among thousands of medical concepts. These connections, for instance, the co-occurrence of diseases and lab-disease correlations can be informative when only a subset of these variables is documented by the clinician. A feasible approach to improving the representation learning of EHR data is to associate relevant medical concepts and utilize these connections. Existing medical ontologies can be the reference for EHR structures, but they place numerous constraints on the data source. Recent progress on graph neural networks (GNN) enables end-to-end learning of topological structures for non-grid or non-sequential data. However, there are problems to be addressed on how to learn the medical graph adaptively and how to understand the effect of the medical graph on representation learning. In this paper, we propose a variationally regularized encoder-decoder graph network that achieves more robustness in graph structure learning by regularizing node representations. Our model outperforms the existing graph and non-graph based methods in various EHR predictive tasks based on both public data and real-world clinical data. Besides the improvements in empirical experiment performances, we provide an interpretation of the effect of variational regularization compared to standard graph neural network, using singular value analysis.

Xiaoyi Zhang, Rodoniki Athanasiadou, Narges Razavian

Twitter data has been shown broadly applicable for public health surveillance. Previous public health studies based on Twitter data have largely relied on keyword-matching or topic models for clustering relevant tweets. However, both methods suffer from the short-length of texts and unpredictable noise that naturally occurs in user-generated contexts. In response, we introduce a deep learning approach that uses hashtags as a form of supervision and learns tweet embeddings for extracting informative textual features. In this case study, we address the specific task of estimating state-level obesity from dietary-related textual features. Our approach yields an estimation that strongly correlates the textual features to government data and outperforms the keyword-matching baseline. The results also demonstrate the potential of discovering risk factors using the textual features. This method is general-purpose and can be applied to a wide range of Twitter-based public health studies.

Josie Williams, Narges Razavian

As machine learning (ML) models, trained on real-world datasets, become common practice, it is critical to measure and quantify their potential biases. In this paper, we focus on renal failure and compare a commonly used traditional risk score, Tangri, with a more powerful machine learning model, which has access to a larger variable set and trained on 1.6 million patients' EHR data. We will compare and discuss the generalization and applicability of these two models, in an attempt to quantify biases of status quo clinical practice, compared to ML-driven models.

Sheng Liu, Chhavi Yadav, Carlos Fernandez-Granda et al.

Early detection is a crucial goal in the study of Alzheimer's Disease (AD). In this work, we describe several techniques to boost the performance of 3D deep convolutional neural networks (CNNs) trained to detect AD using structural brain MRI scans. Specifically, we provide evidence that (1) instance normalization outperforms batch normalization, (2) early spatial downsampling negatively affects performance, (3) widening the model brings consistent gains while increasing the depth does not, and (4) incorporating age information yields moderate improvement. Together, these insights yield an increment of approximately 14% in test accuracy over existing models when distinguishing between patients with AD, mild cognitive impairment, and controls in the ADNI dataset. Similar performance is achieved on an independent dataset.

Jingshu Liu, Zachariah Zhang, Narges Razavian

Early detection of preventable diseases is important for better disease management, improved inter-ventions, and more efficient health-care resource allocation. Various machine learning approacheshave been developed to utilize information in Electronic Health Record (EHR) for this task. Majorityof previous attempts, however, focus on structured fields and lose the vast amount of information inthe unstructured notes. In this work we propose a general multi-task framework for disease onsetprediction that combines both free-text medical notes and structured information. We compareperformance of different deep learning architectures including CNN, LSTM and hierarchical models.In contrast to traditional text-based prediction models, our approach does not require disease specificfeature engineering, and can handle negations and numerical values that exist in the text. Ourresults on a cohort of about 1 million patients show that models using text outperform modelsusing just structured data, and that models capable of using numerical values and negations in thetext, in addition to the raw text, further improve performance. Additionally, we compare differentvisualization methods for medical professionals to interpret model predictions.

Narges Razavian, Jake Marcus, David Sontag

Disparate areas of machine learning have benefited from models that can take raw data with little preprocessing as input and learn rich representations of that raw data in order to perform well on a given prediction task. We evaluate this approach in healthcare by using longitudinal measurements of lab tests, one of the more raw signals of a patient's health state widely available in clinical data, to predict disease onsets. In particular, we train a Long Short-Term Memory (LSTM) recurrent neural network and two novel convolutional neural networks for multi-task prediction of disease onset for 133 conditions based on 18 common lab tests measured over time in a cohort of 298K patients derived from 8 years of administrative claims data. We compare the neural networks to a logistic regression with several hand-engineered, clinically relevant features. We find that the representation-based learning approaches significantly outperform this baseline. We believe that our work suggests a new avenue for patient risk stratification based solely on lab results.

Narges Razavian, David Sontag

Early diagnosis of treatable diseases is essential for improving healthcare, and many diseases' onsets are predictable from annual lab tests and their temporal trends. We introduce a multi-resolution convolutional neural network for early detection of multiple diseases from irregularly measured sparse lab values. Our novel architecture takes as input both an imputed version of the data and a binary observation matrix. For imputing the temporal sparse observations, we develop a flexible, fast to train method for differentiable multivariate kernel regression. Our experiments on data from 298K individuals over 8 years, 18 common lab measurements, and 171 diseases show that the temporal signatures learned via convolution are significantly more predictive than baselines commonly used for early disease diagnosis.