Andrew Y. Ng

Mark Mazumder, Colby Banbury, Xiaozhe Yao et al.

Machine learning research has long focused on models rather than datasets, and prominent datasets are used for common ML tasks without regard to the breadth, difficulty, and faithfulness of the underlying problems. Neglecting the fundamental importance of data has given rise to inaccuracy, bias, and fragility in real-world applications, and research is hindered by saturation across existing dataset benchmarks. In response, we present DataPerf, a community-led benchmark suite for evaluating ML datasets and data-centric algorithms. We aim to foster innovation in data-centric AI through competition, comparability, and reproducibility. We enable the ML community to iterate on datasets, instead of just architectures, and we provide an open, online platform with multiple rounds of challenges to support this iterative development. The first iteration of DataPerf contains five benchmarks covering a wide spectrum of data-centric techniques, tasks, and modalities in vision, speech, acquisition, debugging, and diffusion prompting, and we support hosting new contributed benchmarks from the community. The benchmarks, online evaluation platform, and baseline implementations are open source, and the MLCommons Association will maintain DataPerf to ensure long-term benefits to academia and industry.

Elliott Watkiss-Leek, Reham Alharbi, Harry Rostron et al.

Competency question (CQ) elicitation represents a critical but resource-intensive bottleneck in ontology engineering. This foundational phase is often hampered by the communication gap between domain experts, who possess the necessary knowledge, and ontology engineers, who formalise it. This paper introduces IDEA2, a novel, semi-automated workflow that integrates Large Language Models (LLMs) within a collaborative, expert-in-the-loop process to address this challenge. The methodology is characterised by a core iterative loop: an initial LLM-based extraction of CQs from requirement documents, a co-creational review and feedback phase by domain experts on an accessible collaborative platform, and an iterative, feedback-driven reformulation of rejected CQs by an LLM until consensus is achieved. To ensure transparency and reproducibility, the entire lifecycle of each CQ is tracked using a provenance model that captures the full lineage of edits, anonymised feedback, and generation parameters. The workflow was validated in 2 real-world scenarios (scientific data, cultural heritage), demonstrating that IDEA2 can accelerate the requirements engineering process, improve the acceptance and relevance of the resulting CQs, and exhibit high usability and effectiveness among domain experts. We release all code and experiments at https://github.com/KE-UniLiv/IDEA2

Bryan Zhu, Nicholas Lui, Jeremy Irvin et al.

Reducing methane emissions is essential for mitigating global warming. To attribute methane emissions to their sources, a comprehensive dataset of methane source infrastructure is necessary. Recent advancements with deep learning on remotely sensed imagery have the potential to identify the locations and characteristics of methane sources, but there is a substantial lack of publicly available data to enable machine learning researchers and practitioners to build automated mapping approaches. To help fill this gap, we construct a multi-sensor dataset called METER-ML containing 86,599 georeferenced NAIP, Sentinel-1, and Sentinel-2 images in the U.S. labeled for the presence or absence of methane source facilities including concentrated animal feeding operations, coal mines, landfills, natural gas processing plants, oil refineries and petroleum terminals, and wastewater treatment plants. We experiment with a variety of models that leverage different spatial resolutions, spatial footprints, image products, and spectral bands. We find that our best model achieves an area under the precision recall curve of 0.915 for identifying concentrated animal feeding operations and 0.821 for oil refineries and petroleum terminals on an expert-labeled test set, suggesting the potential for large-scale mapping. We make METER-ML freely available at https://stanfordmlgroup.github.io/projects/meter-ml/ to support future work on automated methane source mapping.

Jeremy Andrew Irvin, Jiaqi Han, Zikui Wang et al.

Generative models have the potential to transform the way we emulate Earth's changing climate. Previous generative approaches rely on weather-scale autoregression for climate emulation, but this is inherently slow for long climate horizons and has yet to demonstrate stable rollouts under nonstationary forcings. Here, we introduce Spatiotemporal Pyramid Flows (SPF), a new class of flow matching approaches that model data hierarchically across spatial and temporal scales. Inspired by cascaded video models, SPF partitions the generative trajectory into a spatiotemporal pyramid, progressively increasing spatial resolution to reduce computation and coupling each stage with an associated timescale to enable direct sampling at any temporal level in the pyramid. This design, together with conditioning each stage on prescribed physical forcings (e.g., greenhouse gases or aerosols), enables efficient, parallel climate emulation at multiple timescales. On ClimateBench, SPF outperforms strong flow matching baselines and pre-trained models at yearly and monthly timescales while offering fast sampling, especially at coarser temporal levels. To scale SPF, we curate ClimateSuite, the largest collection of Earth system simulations to date, comprising over 33,000 simulation-years across ten climate models and the first dataset to include simulations of climate interventions. We find that the scaled SPF model demonstrates good generalization to held-out scenarios across climate models. Together, SPF and ClimateSuite provide a foundation for accurate, efficient, probabilistic climate emulation across temporal scales and realistic future scenarios. Data and code is publicly available at https://github.com/stanfordmlgroup/spf .

Vivek Shankar, Xiaoli Yang, Vrishab Krishna et al.

The accurate classification of lymphoma subtypes using hematoxylin and eosin (H&E)-stained tissue is complicated by the wide range of morphological features these cancers can exhibit. We present LymphoML - an interpretable machine learning method that identifies morphologic features that correlate with lymphoma subtypes. Our method applies steps to process H&E-stained tissue microarray cores, segment nuclei and cells, compute features encompassing morphology, texture, and architecture, and train gradient-boosted models to make diagnostic predictions. LymphoML's interpretable models, developed on a limited volume of H&E-stained tissue, achieve non-inferior diagnostic accuracy to pathologists using whole-slide images and outperform black box deep-learning on a dataset of 670 cases from Guatemala spanning 8 lymphoma subtypes. Using SHapley Additive exPlanation (SHAP) analysis, we assess the impact of each feature on model prediction and find that nuclear shape features are most discriminative for DLBCL (F1-score: 78.7%) and classical Hodgkin lymphoma (F1-score: 74.5%). Finally, we provide the first demonstration that a model combining features from H&E-stained tissue with features from a standardized panel of 6 immunostains results in a similar diagnostic accuracy (85.3%) to a 46-stain panel (86.1%).

Yi-Lin Tsai, Jeremy Irvin, Suhas Chundi et al.

Taiwan has the highest susceptibility to and fatalities from debris flows worldwide. The existing debris flow warning system in Taiwan, which uses a time-weighted measure of rainfall, leads to alerts when the measure exceeds a predefined threshold. However, this system generates many false alarms and misses a substantial fraction of the actual debris flows. Towards improving this system, we implemented five machine learning models that input historical rainfall data and predict whether a debris flow will occur within a selected time. We found that a random forest model performed the best among the five models and outperformed the existing system in Taiwan. Furthermore, we identified the rainfall trajectories strongly related to debris flow occurrences and explored trade-offs between the risks of missing debris flows versus frequent false alerts. These results suggest the potential for machine learning models trained on hourly rainfall data alone to save lives while reducing false alerts.

Tianyuan Huang, Timothy Dai, Zhecheng Wang et al.

Neighborhood gentrification plays a significant role in shaping the social and economic well-being of both individuals and communities at large. While some efforts have been made to detect gentrification in cities, existing approaches rely mainly on estimated measures from survey data, require substantial work of human labeling, and are limited in characterizing the neighborhood as a whole. We propose a novel approach to detecting neighborhood gentrification at a large-scale based on the physical appearance of neighborhoods by incorporating historical street-level visual data. We show the effectiveness of the proposed method by comparing results from our approach with gentrification measures from previous literature and case studies. Our approach has the potential to supplement existing indicators of gentrification and become a valid resource for urban researchers and policy makers.

Ji Hun Wang, Jeremy Irvin, Beri Kohen Behar et al.

Deep learning for detecting objects in remotely sensed imagery can enable new technologies for important applications including mitigating climate change. However, these models often require large datasets labeled with bounding box annotations which are expensive to curate, prohibiting the development of models for new tasks and geographies. To address this challenge, we develop weakly-semi-supervised object detection (WSSOD) models on remotely sensed imagery which can leverage a small amount of bounding boxes together with a large amount of point labels that are easy to acquire at scale in geospatial data. We train WSSOD models which use large amounts of point-labeled images with varying fractions of bounding box labeled images in FAIR1M and a wind turbine detection dataset, and demonstrate that they substantially outperform fully supervised models trained with the same amount of bounding box labeled images on both datasets. Furthermore, we find that the WSSOD models trained with 2-10x fewer bounding box labeled images can perform similarly to or outperform fully supervised models trained on the full set of bounding-box labeled images. We believe that the approach can be extended to other remote sensing tasks to reduce reliance on bounding box labels and increase development of models for impactful applications.

Yixing Jiang, Jeremy Irvin, Ji Hun Wang et al.

Large language models are effective at few-shot in-context learning (ICL). Recent advancements in multimodal foundation models have enabled unprecedentedly long context windows, presenting an opportunity to explore their capability to perform ICL with many more demonstrating examples. In this work, we evaluate the performance of multimodal foundation models scaling from few-shot to many-shot ICL. We benchmark GPT-4o and Gemini 1.5 Pro across 14 datasets spanning multiple domains (natural imagery, medical imagery, remote sensing, and molecular imagery) and tasks (image classification, visual QA, and object localization). We observe that many-shot ICL, including up to almost 2,000 demonstrating examples, leads to substantial improvements compared to few-shot (<100 examples) ICL across all of the datasets. Further, Gemini 1.5 Pro performance continues to improve log-linearly up to the maximum number of tested examples on many datasets. We also find open-weights multimodal foundation models like Llama 3.2-Vision do not benefit from the demonstrating examples, highlighting an important gap between open and closed multimodal foundation models. Given the high inference costs required for many-shot ICL, we also explore the impact of batching multiple queries in a single API call. We show that batching up to 50 queries can lead to performance improvements under zero-shot and many-shot ICL, with substantial gains in the zero-shot setting on multiple datasets, while drastically reducing per-query cost and latency. Finally, while GPT-4o and Gemini 1.5 Pro achieve similar zero-shot performance across the datasets, Gemini 1.5 Pro learns more quickly than GPT-4o on most datasets. Our results suggest that many-shot ICL could enable users to efficiently adapt multimodal foundation models to new applications and domains. Our codebase is publicly available at https://github.com/stanfordmlgroup/ManyICL .

Yixing Jiang, Kameron C. Black, Gloria Geng et al.

Recent large language models (LLMs) have demonstrated significant advancements, particularly in their ability to serve as agents thereby surpassing their traditional role as chatbots. These agents can leverage their planning and tool utilization capabilities to address tasks specified at a high level. However, a standardized dataset to benchmark the agent capabilities of LLMs in medical applications is currently lacking, making the evaluation of LLMs on complex tasks in interactive healthcare environments challenging. To address this gap, we introduce MedAgentBench, a broad evaluation suite designed to assess the agent capabilities of large language models within medical records contexts. MedAgentBench encompasses 300 patient-specific clinically-derived tasks from 10 categories written by human physicians, realistic profiles of 100 patients with over 700,000 data elements, a FHIR-compliant interactive environment, and an accompanying codebase. The environment uses the standard APIs and communication infrastructure used in modern EMR systems, so it can be easily migrated into live EMR systems. MedAgentBench presents an unsaturated agent-oriented benchmark that current state-of-the-art LLMs exhibit some ability to succeed at. The best model (Claude 3.5 Sonnet v2) achieves a success rate of 69.67%. However, there is still substantial space for improvement which gives the community a next direction to optimize. Furthermore, there is significant variation in performance across task categories. MedAgentBench establishes this and is publicly available at https://github.com/stanfordmlgroup/MedAgentBench , offering a valuable framework for model developers to track progress and drive continuous improvements in the agent capabilities of large language models within the medical domain.

Tanvi Deshpande, Eva Prakash, Elsie Gyang Ross et al.

The high cost of creating pixel-by-pixel gold-standard labels, limited expert availability, and presence of diverse tasks make it challenging to generate segmentation labels to train deep learning models for medical imaging tasks. In this work, we present a new approach to overcome the hurdle of costly medical image labeling by leveraging foundation models like Segment Anything Model (SAM) and its medical alternate MedSAM. Our pipeline has the ability to generate weak labels for any unlabeled medical image and subsequently use it to augment label-scarce datasets. We perform this by leveraging a model trained on a few gold-standard labels and using it to intelligently prompt MedSAM for weak label generation. This automation eliminates the manual prompting step in MedSAM, creating a streamlined process for generating labels for both real and synthetic images, regardless of quantity. We conduct experiments on label-scarce settings for multiple tasks pertaining to modalities ranging from ultrasound, dermatology, and X-rays to demonstrate the usefulness of our pipeline. The code is available at https://github.com/stanfordmlgroup/Auto-Generate-WLs/.

Sharon Zhou, Eric Zelikman, Fred Lu et al.

Learning disentangled representations is regarded as a fundamental task for improving the generalization, robustness, and interpretability of generative models. However, measuring disentanglement has been challenging and inconsistent, often dependent on an ad-hoc external model or specific to a certain dataset. To address this, we present a method for quantifying disentanglement that only uses the generative model, by measuring the topological similarity of conditional submanifolds in the learned representation. This method showcases both unsupervised and supervised variants. To illustrate the effectiveness and applicability of our method, we empirically evaluate several state-of-the-art models across multiple datasets. We find that our method ranks models similarly to existing methods. We make ourcode publicly available at https://github.com/stanfordmlgroup/disentanglement.

Tony Duan, Anand Avati, Daisy Yi Ding et al.

We present Natural Gradient Boosting (NGBoost), an algorithm for generic probabilistic prediction via gradient boosting. Typical regression models return a point estimate, conditional on covariates, but probabilistic regression models output a full probability distribution over the outcome space, conditional on the covariates. This allows for predictive uncertainty estimation -- crucial in applications like healthcare and weather forecasting. NGBoost generalizes gradient boosting to probabilistic regression by treating the parameters of the conditional distribution as targets for a multiparameter boosting algorithm. Furthermore, we show how the Natural Gradient is required to correct the training dynamics of our multiparameter boosting approach. NGBoost can be used with any base learner, any family of distributions with continuous parameters, and any scoring rule. NGBoost matches or exceeds the performance of existing methods for probabilistic prediction while offering additional benefits in flexibility, scalability, and usability. An open-source implementation is available at github.com/stanfordmlgroup/ngboost.

Chih-Ying Liu, Jeya Maria Jose Valanarasu, Camila Gonzalez et al.

Most deep learning models in medical imaging are trained on adult data with unclear performance on pediatric images. In this work, we aim to address this challenge in the context of automated anatomy segmentation in whole-body Computed Tomography (CT). We evaluate the performance of CT organ segmentation algorithms trained on adult data when applied to pediatric CT volumes and identify substantial age-dependent underperformance. We subsequently propose and evaluate strategies, including data augmentation and continual learning approaches, to achieve good segmentation accuracy across all age groups. Our best-performing model, trained using continual learning, achieves high segmentation accuracy on both adult and pediatric data (Dice scores of 0.90 and 0.84 respectively).

Eva Prakash, Jeya Maria Jose Valanarasu, Zhihong Chen et al.

Purpose: To explore best-practice approaches for generating synthetic chest X-ray images and augmenting medical imaging datasets to optimize the performance of deep learning models in downstream tasks like classification and segmentation. Materials and Methods: We utilized a latent diffusion model to condition the generation of synthetic chest X-rays on text prompts and/or segmentation masks. We explored methods like using a proxy model and using radiologist feedback to improve the quality of synthetic data. These synthetic images were then generated from relevant disease information or geometrically transformed segmentation masks and added to ground truth training set images from the CheXpert, CANDID-PTX, SIIM, and RSNA Pneumonia datasets to measure improvements in classification and segmentation model performance on the test sets. F1 and Dice scores were used to evaluate classification and segmentation respectively. One-tailed t-tests with Bonferroni correction assessed the statistical significance of performance improvements with synthetic data. Results: Across all experiments, the synthetic data we generated resulted in a maximum mean classification F1 score improvement of 0.150453 (CI: 0.099108-0.201798; P=0.0031) compared to using only real data. For segmentation, the maximum Dice score improvement was 0.14575 (CI: 0.108267-0.183233; P=0.0064). Conclusion: Best practices for generating synthetic chest X-ray images for downstream tasks include conditioning on single-disease labels or geometrically transformed segmentation masks, as well as potentially using proxy modeling for fine-tuning such generations.

Muhammad Ahmed Chaudhry, Lyna Kim, Jeremy Irvin et al.

Clouds play a significant role in global temperature regulation through their effect on planetary albedo. Anthropogenic emissions of aerosols can alter the albedo of clouds, but the extent of this effect, and its consequent impact on temperature change, remains uncertain. Human-induced clouds caused by ship aerosol emissions, commonly referred to as ship tracks, provide visible manifestations of this effect distinct from adjacent cloud regions and therefore serve as a useful sandbox to study human-induced clouds. However, the lack of large-scale ship track data makes it difficult to deduce their general effects on cloud formation. Towards developing automated approaches to localize ship tracks at scale, we present CloudTracks, a dataset containing 3,560 satellite images labeled with more than 12,000 ship track instance annotations. We train semantic segmentation and instance segmentation model baselines on our dataset and find that our best model substantially outperforms previous state-of-the-art for ship track localization (61.29 vs. 48.65 IoU). We also find that the best instance segmentation model is able to identify the number of ship tracks in each image more accurately than the previous state-of-the-art (1.64 vs. 4.99 MAE). However, we identify cases where the best model struggles to accurately localize and count ship tracks, so we believe CloudTracks will stimulate novel machine learning approaches to better detect elongated and overlapping features in satellite images. We release our dataset openly at {zenodo.org/records/10042922}.

Cara Van Uden, Jeremy Irvin, Mars Huang et al.

Self-supervised learning (SSL) enables label efficient training for machine learning models. This is essential for domains such as medical imaging, where labels are costly and time-consuming to curate. However, the most effective supervised or SSL strategy for transferring models to different healthcare systems or novel tasks is not well understood. In this work, we systematically experiment with a variety of supervised and self-supervised pretraining strategies using multimodal datasets of medical images (chest X-rays) and text (radiology reports). We then evaluate their performance on data from two external institutions with diverse sets of tasks. In addition, we experiment with different transfer learning strategies to effectively adapt these pretrained models to new tasks and healthcare systems. Our empirical results suggest that multimodal SSL gives substantial gains over unimodal SSL in performance across new healthcare systems and tasks, comparable to models pretrained with full supervision. We demonstrate additional performance gains with models further adapted to the new dataset and task, using multimodal domain-adaptive pretraining (DAPT), linear probing then finetuning (LP-FT), and both methods combined. We offer suggestions for alternative models to use in scenarios where not all of these additions are feasible. Our results provide guidance for improving the generalization of medical image interpretation models to new healthcare systems and novel tasks.

Jon Braatz, Pranav Rajpurkar, Stephanie Zhang et al.

In recent years, deep learning has successfully been applied to automate a wide variety of tasks in diagnostic histopathology. However, fast and reliable localization of small-scale regions-of-interest (ROI) has remained a key challenge, as discriminative morphologic features often occupy only a small fraction of a gigapixel-scale whole-slide image (WSI). In this paper, we propose a sparse WSI analysis method for the rapid identification of high-power ROI for WSI-level classification. We develop an evaluation framework inspired by the early classification literature, in order to quantify the tradeoff between diagnostic performance and inference time for sparse analytic approaches. We test our method on a common but time-consuming task in pathology - that of diagnosing gastric intestinal metaplasia (GIM) on hematoxylin and eosin (H&E)-stained slides from endoscopic biopsy specimens. GIM is a well-known precursor lesion along the pathway to development of gastric cancer. We performed a thorough evaluation of the performance and inference time of our approach on a test set of GIM-positive and GIM-negative WSI, finding that our method successfully detects GIM in all positive WSI, with a WSI-level classification area under the receiver operating characteristic curve (AUC) of 0.98 and an average precision (AP) of 0.95. Furthermore, we show that our method can attain these metrics in under one minute on a standard CPU. Our results are applicable toward the goal of developing neural networks that can easily be deployed in clinical settings to support pathologists in quickly localizing and diagnosing small-scale morphologic features in WSI.

Cécile Logé, Emily Ross, David Yaw Amoah Dadey et al.

Recent advances in Natural Language Processing (NLP), and specifically automated Question Answering (QA) systems, have demonstrated both impressive linguistic fluency and a pernicious tendency to reflect social biases. In this study, we introduce Q-Pain, a dataset for assessing bias in medical QA in the context of pain management, one of the most challenging forms of clinical decision-making. Along with the dataset, we propose a new, rigorous framework, including a sample experimental design, to measure the potential biases present when making treatment decisions. We demonstrate its use by assessing two reference Question-Answering systems, GPT-2 and GPT-3, and find statistically significant differences in treatment between intersectional race-gender subgroups, thus reaffirming the risks posed by AI in medical settings, and the need for datasets like ours to ensure safety before medical AI applications are deployed.

Saahil Jain, Ashwin Agrawal, Adriel Saporta et al.

Extracting structured clinical information from free-text radiology reports can enable the use of radiology report information for a variety of critical healthcare applications. In our work, we present RadGraph, a dataset of entities and relations in full-text chest X-ray radiology reports based on a novel information extraction schema we designed to structure radiology reports. We release a development dataset, which contains board-certified radiologist annotations for 500 radiology reports from the MIMIC-CXR dataset (14,579 entities and 10,889 relations), and a test dataset, which contains two independent sets of board-certified radiologist annotations for 100 radiology reports split equally across the MIMIC-CXR and CheXpert datasets. Using these datasets, we train and test a deep learning model, RadGraph Benchmark, that achieves a micro F1 of 0.82 and 0.73 on relation extraction on the MIMIC-CXR and CheXpert test sets respectively. Additionally, we release an inference dataset, which contains annotations automatically generated by RadGraph Benchmark across 220,763 MIMIC-CXR reports (around 6 million entities and 4 million relations) and 500 CheXpert reports (13,783 entities and 9,908 relations) with mappings to associated chest radiographs. Our freely available dataset can facilitate a wide range of research in medical natural language processing, as well as computer vision and multi-modal learning when linked to chest radiographs.

Tianyuan Huang, Zhecheng Wang, Hao Sheng et al.

Recent urbanization has coincided with the enrichment of geotagged data, such as street view and point-of-interest (POI). Region embedding enhanced by the richer data modalities has enabled researchers and city administrators to understand the built environment, socioeconomics, and the dynamics of cities better. While some efforts have been made to simultaneously use multi-modal inputs, existing methods can be improved by incorporating different measures of 'proximity' in the same embedding space - leveraging not only the data that characterizes the regions (e.g., street view, local businesses pattern) but also those that depict the relationship between regions (e.g., trips, road network). To this end, we propose a novel approach to integrate multi-modal geotagged inputs as either node or edge features of a multi-graph based on their relations with the neighborhood region (e.g., tiles, census block, ZIP code region, etc.). We then learn the neighborhood representation based on a contrastive-sampling scheme from the multi-graph. Specifically, we use street view images and POI features to characterize neighborhoods (nodes) and use human mobility to characterize the relationship between neighborhoods (directed edges). We show the effectiveness of the proposed methods with quantitative downstream tasks as well as qualitative analysis of the embedding space: The embedding we trained outperforms the ones using only unimodal data as regional inputs.

Bryan Gopal, Ryan W. Han, Gautham Raghupathi et al.

We propose 3KG, a physiologically-inspired contrastive learning approach that generates views using 3D augmentations of the 12-lead electrocardiogram. We evaluate representation quality by fine-tuning a linear layer for the downstream task of 23-class diagnosis on the PhysioNet 2020 challenge training data and find that 3KG achieves a $9.1\%$ increase in mean AUC over the best self-supervised baseline when trained on $1\%$ of labeled data. Our empirical analysis shows that combining spatial and temporal augmentations produces the strongest representations. In addition, we investigate the effect of this physiologically-inspired pretraining on downstream performance on different disease subgroups and find that 3KG makes the greatest gains for conduction and rhythm abnormalities. Our method allows for flexibility in incorporating other self-supervised strategies and highlights the potential for similar modality-specific augmentations for other biomedical signals.

Saahil Jain, Akshay Smit, Andrew Y. Ng et al.

Although deep learning models for chest X-ray interpretation are commonly trained on labels generated by automatic radiology report labelers, the impact of improvements in report labeling on the performance of chest X-ray classification models has not been systematically investigated. We first compare the CheXpert, CheXbert, and VisualCheXbert labelers on the task of extracting accurate chest X-ray image labels from radiology reports, reporting that the VisualCheXbert labeler outperforms the CheXpert and CheXbert labelers. Next, after training image classification models using labels generated from the different radiology report labelers on one of the largest datasets of chest X-rays, we show that an image classification model trained on labels from the VisualCheXbert labeler outperforms image classification models trained on labels from the CheXpert and CheXbert labelers. Our work suggests that recent improvements in radiology report labeling can translate to the development of higher performing chest X-ray classification models.

Akshay Smit, Damir Vrabac, Yujie He et al.

We propose a selective learning method using meta-learning and deep reinforcement learning for medical image interpretation in the setting of limited labeling resources. Our method, MedSelect, consists of a trainable deep learning selector that uses image embeddings obtained from contrastive pretraining for determining which images to label, and a non-parametric selector that uses cosine similarity to classify unseen images. We demonstrate that MedSelect learns an effective selection strategy outperforming baseline selection strategies across seen and unseen medical conditions for chest X-ray interpretation. We also perform an analysis of the selections performed by MedSelect comparing the distribution of latent embeddings and clinical features, and find significant differences compared to the strongest performing baseline. We believe that our method may be broadly applicable across medical imaging settings where labels are expensive to acquire.

Emma Chen, Andy Kim, Rayan Krishnan et al.

A major obstacle to the integration of deep learning models for chest x-ray interpretation into clinical settings is the lack of understanding of their failure modes. In this work, we first investigate whether there are patient subgroups that chest x-ray models are likely to misclassify. We find that patient age and the radiographic finding of lung lesion, pneumothorax or support devices are statistically relevant features for predicting misclassification for some chest x-ray models. Second, we develop misclassification predictors on chest x-ray models using their outputs and clinical features. We find that our best performing misclassification identifier achieves an AUROC close to 0.9 for most diseases. Third, employing our misclassification identifiers, we develop a corrective algorithm to selectively flip model predictions that have high likelihood of misclassification at inference time. We observe F1 improvement on the prediction of Consolidation (0.008 [95% CI 0.005, 0.010]) and Edema (0.003, [95% CI 0.001, 0.006]). By carrying out our investigation on ten distinct and high-performing chest x-ray models, we are able to derive insights across model architectures and offer a generalizable framework applicable to other medical imaging tasks.

Siyu Shi, Ishaan Malhi, Kevin Tran et al.

We systematically evaluate the performance of deep learning models in the presence of diseases not labeled for or present during training. First, we evaluate whether deep learning models trained on a subset of diseases (seen diseases) can detect the presence of any one of a larger set of diseases. We find that models tend to falsely classify diseases outside of the subset (unseen diseases) as "no disease". Second, we evaluate whether models trained on seen diseases can detect seen diseases when co-occurring with diseases outside the subset (unseen diseases). We find that models are still able to detect seen diseases even when co-occurring with unseen diseases. Third, we evaluate whether feature representations learned by models may be used to detect the presence of unseen diseases given a small labeled set of unseen diseases. We find that the penultimate layer of the deep neural network provides useful features for unseen disease detection. Our results can inform the safe clinical deployment of deep learning models trained on a non-exhaustive set of disease classes.

Saahil Jain, Akshay Smit, Steven QH Truong et al.

Automatic extraction of medical conditions from free-text radiology reports is critical for supervising computer vision models to interpret medical images. In this work, we show that radiologists labeling reports significantly disagree with radiologists labeling corresponding chest X-ray images, which reduces the quality of report labels as proxies for image labels. We develop and evaluate methods to produce labels from radiology reports that have better agreement with radiologists labeling images. Our best performing method, called VisualCheXbert, uses a biomedically-pretrained BERT model to directly map from a radiology report to the image labels, with a supervisory signal determined by a computer vision model trained to detect medical conditions from chest X-ray images. We find that VisualCheXbert outperforms an approach using an existing radiology report labeler by an average F1 score of 0.14 (95% CI 0.12, 0.17). We also find that VisualCheXbert better agrees with radiologists labeling chest X-ray images than do radiologists labeling the corresponding radiology reports by an average F1 score across several medical conditions of between 0.12 (95% CI 0.09, 0.15) and 0.21 (95% CI 0.18, 0.24).

Yen Nhi Truong Vu, Richard Wang, Niranjan Balachandar et al.

Self-supervised contrastive learning between pairs of multiple views of the same image has been shown to successfully leverage unlabeled data to produce meaningful visual representations for both natural and medical images. However, there has been limited work on determining how to select pairs for medical images, where availability of patient metadata can be leveraged to improve representations. In this work, we develop a method to select positive pairs coming from views of possibly different images through the use of patient metadata. We compare strategies for selecting positive pairs for chest X-ray interpretation including requiring them to be from the same patient, imaging study or laterality. We evaluate downstream task performance by fine-tuning the linear layer on 1% of the labeled dataset for pleural effusion classification. Our best performing positive pair selection strategy, which involves using images from the same patient from the same study across all lateralities, achieves a performance increase of 14.4% in mean AUC from the ImageNet pretrained baseline. Our controlled experiments show that the keys to improving downstream performance on disease classification are (1) using patient metadata to appropriately create positive pairs from different images with the same underlying pathologies, and (2) maximizing the number of different images used in query pairing. In addition, we explore leveraging patient metadata to select hard negative pairs for contrastive learning, but do not find improvement over baselines that do not use metadata. Our method is broadly applicable to medical image interpretation and allows flexibility for incorporating medical insights in choosing pairs for contrastive learning.

Soham Gadgil, Mark Endo, Emily Wen et al.

Medical image segmentation models are typically supervised by expert annotations at the pixel-level, which can be expensive to acquire. In this work, we propose a method that combines the high quality of pixel-level expert annotations with the scale of coarse DNN-generated saliency maps for training multi-label semantic segmentation models. We demonstrate the application of our semi-supervised method, which we call CheXseg, on multi-label chest X-ray interpretation. We find that CheXseg improves upon the performance (mIoU) of fully-supervised methods that use only pixel-level expert annotations by 9.7% and weakly-supervised methods that use only DNN-generated saliency maps by 73.1%. Our best method is able to match radiologist agreement on three out of ten pathologies and reduces the overall performance gap by 57.2% as compared to weakly-supervised methods.

Pranav Rajpurkar, Anirudh Joshi, Anuj Pareek et al.

Recent advances in training deep learning models have demonstrated the potential to provide accurate chest X-ray interpretation and increase access to radiology expertise. However, poor generalization due to data distribution shifts in clinical settings is a key barrier to implementation. In this study, we measured the diagnostic performance for 8 different chest X-ray models when applied to (1) smartphone photos of chest X-rays and (2) external datasets without any finetuning. All models were developed by different groups and submitted to the CheXpert challenge, and re-applied to test datasets without further tuning. We found that (1) on photos of chest X-rays, all 8 models experienced a statistically significant drop in task performance, but only 3 performed significantly worse than radiologists on average, and (2) on the external set, none of the models performed statistically significantly worse than radiologists, and five models performed statistically significantly better than radiologists. Our results demonstrate that some chest X-ray models, under clinically relevant distribution shifts, were comparable to radiologists while other models were not. Future work should investigate aspects of model training procedures and dataset collection that influence generalization in the presence of data distribution shifts.

Alexander Ke, William Ellsworth, Oishi Banerjee et al.

Deep learning methods for chest X-ray interpretation typically rely on pretrained models developed for ImageNet. This paradigm assumes that better ImageNet architectures perform better on chest X-ray tasks and that ImageNet-pretrained weights provide a performance boost over random initialization. In this work, we compare the transfer performance and parameter efficiency of 16 popular convolutional architectures on a large chest X-ray dataset (CheXpert) to investigate these assumptions. First, we find no relationship between ImageNet performance and CheXpert performance for both models without pretraining and models with pretraining. Second, we find that, for models without pretraining, the choice of model family influences performance more than size within a family for medical imaging tasks. Third, we observe that ImageNet pretraining yields a statistically significant boost in performance across architectures, with a higher boost for smaller architectures. Fourth, we examine whether ImageNet architectures are unnecessarily large for CheXpert by truncating final blocks from pretrained models, and find that we can make models 3.25x more parameter-efficient on average without a statistically significant drop in performance. Our work contributes new experimental evidence about the relation of ImageNet to chest x-ray interpretation performance.

Hao Sheng, Jeremy Irvin, Sasankh Munukutla et al.

At least a quarter of the warming that the Earth is experiencing today is due to anthropogenic methane emissions. There are multiple satellites in orbit and planned for launch in the next few years which can detect and quantify these emissions; however, to attribute methane emissions to their sources on the ground, a comprehensive database of the locations and characteristics of emission sources worldwide is essential. In this work, we develop deep learning algorithms that leverage freely available high-resolution aerial imagery to automatically detect oil and gas infrastructure, one of the largest contributors to global methane emissions. We use the best algorithm, which we call OGNet, together with expert review to identify the locations of oil refineries and petroleum terminals in the U.S. We show that OGNet detects many facilities which are not present in four standard public datasets of oil and gas infrastructure. All detected facilities are associated with characteristics known to contribute to methane emissions, including the infrastructure type and the number of storage tanks. The data curated and produced in this study is freely available at http://stanfordmlgroup.github.io/projects/ognet .

Pranav Rajpurkar, Anirudh Joshi, Anuj Pareek et al.

The use of smartphones to take photographs of chest x-rays represents an appealing solution for scaled deployment of deep learning models for chest x-ray interpretation. However, the performance of chest x-ray algorithms on photos of chest x-rays has not been thoroughly investigated. In this study, we measured the diagnostic performance for 8 different chest x-ray models when applied to photos of chest x-rays. All models were developed by different groups and submitted to the CheXpert challenge, and re-applied to smartphone photos of x-rays in the CheXphoto dataset without further tuning. We found that several models had a drop in performance when applied to photos of chest x-rays, but even with this drop, some models still performed comparably to radiologists. Further investigation could be directed towards understanding how different model training procedures may affect model generalization to photos of chest x-rays.

Jeremy Irvin, Hao Sheng, Neel Ramachandran et al.

Characterizing the processes leading to deforestation is critical to the development and implementation of targeted forest conservation and management policies. In this work, we develop a deep learning model called ForestNet to classify the drivers of primary forest loss in Indonesia, a country with one of the highest deforestation rates in the world. Using satellite imagery, ForestNet identifies the direct drivers of deforestation in forest loss patches of any size. We curate a dataset of Landsat 8 satellite images of known forest loss events paired with driver annotations from expert interpreters. We use the dataset to train and validate the models and demonstrate that ForestNet substantially outperforms other standard driver classification approaches. In order to support future research on automated approaches to deforestation driver classification, the dataset curated in this study is publicly available at https://stanfordmlgroup.github.io/projects/forestnet .

Viswesh Krishna, Anirudh Joshi, Philip L. Bulterys et al.

The application of deep learning to pathology assumes the existence of digital whole slide images of pathology slides. However, slide digitization is bottlenecked by the high cost of precise motor stages in slide scanners that are needed for position information used for slide stitching. We propose GloFlow, a two-stage method for creating a whole slide image using optical flow-based image registration with global alignment using a computationally tractable graph-pruning approach. In the first stage, we train an optical flow predictor to predict pairwise translations between successive video frames to approximate a stitch. In the second stage, this approximate stitch is used to create a neighborhood graph to produce a corrected stitch. On a simulated dataset of video scans of WSIs, we find that our method outperforms known approaches to slide-stitching, and stitches WSIs resembling those produced by slide scanners.

Hari Sowrirajan, Jingbo Yang, Andrew Y. Ng et al.

Contrastive learning is a form of self-supervision that can leverage unlabeled data to produce pretrained models. While contrastive learning has demonstrated promising results on natural image classification tasks, its application to medical imaging tasks like chest X-ray interpretation has been limited. In this work, we propose MoCo-CXR, which is an adaptation of the contrastive learning method Momentum Contrast (MoCo), to produce models with better representations and initializations for the detection of pathologies in chest X-rays. In detecting pleural effusion, we find that linear models trained on MoCo-CXR-pretrained representations outperform those without MoCo-CXR-pretrained representations, indicating that MoCo-CXR-pretrained representations are of higher-quality. End-to-end fine-tuning experiments reveal that a model initialized via MoCo-CXR-pretraining outperforms its non-MoCo-CXR-pretrained counterpart. We find that MoCo-CXR-pretraining provides the most benefit with limited labeled training data. Finally, we demonstrate similar results on a target Tuberculosis dataset unseen during pretraining, indicating that MoCo-CXR-pretraining endows models with representations and transferability that can be applied across chest X-ray datasets and tasks.

Eric Zelikman, Sharon Zhou, Jeremy Irvin et al.

Advancing probabilistic solar forecasting methods is essential to supporting the integration of solar energy into the electricity grid. In this work, we develop a variety of state-of-the-art probabilistic models for forecasting solar irradiance. We investigate the use of post-hoc calibration techniques for ensuring well-calibrated probabilistic predictions. We train and evaluate the models using public data from seven stations in the SURFRAD network, and demonstrate that the best model, NGBoost, achieves higher performance at an intra-hourly resolution than the best benchmark solar irradiance forecasting model across all stations. Further, we show that NGBoost with CRUDE post-hoc calibration achieves comparable performance to a numerical weather prediction model on hourly-resolution forecasting.

Damir Vrabac, Akshay Smit, Rebecca Rojansky et al.

Diffuse Large B-Cell Lymphoma (DLBCL) is the most common non-Hodgkin lymphoma. Though histologically DLBCL shows varying morphologies, no morphologic features have been consistently demonstrated to correlate with prognosis. We present a morphologic analysis of histology sections from 209 DLBCL cases with associated clinical and cytogenetic data. Duplicate tissue core sections were arranged in tissue microarrays (TMAs), and replicate sections were stained with H&E and immunohistochemical stains for CD10, BCL6, MUM1, BCL2, and MYC. The TMAs are accompanied by pathologist-annotated regions-of-interest (ROIs) that identify areas of tissue representative of DLBCL. We used a deep learning model to segment all tumor nuclei in the ROIs, and computed several geometric features for each segmented nucleus. We fit a Cox proportional hazards model to demonstrate the utility of these geometric features in predicting survival outcome, and found that it achieved a C-index (95% CI) of 0.635 (0.574,0.691). Our finding suggests that geometric features computed from tumor nuclei are of prognostic importance, and should be validated in prospective studies.

Nick A. Phillips, Pranav Rajpurkar, Mark Sabini et al.

Clinical deployment of deep learning algorithms for chest x-ray interpretation requires a solution that can integrate into the vast spectrum of clinical workflows across the world. An appealing approach to scaled deployment is to leverage the ubiquity of smartphones by capturing photos of x-rays to share with clinicians using messaging services like WhatsApp. However, the application of chest x-ray algorithms to photos of chest x-rays requires reliable classification in the presence of artifacts not typically encountered in digital x-rays used to train machine learning models. We introduce CheXphoto, a dataset of smartphone photos and synthetic photographic transformations of chest x-rays sampled from the CheXpert dataset. To generate CheXphoto we (1) automatically and manually captured photos of digital x-rays under different settings, and (2) generated synthetic transformations of digital x-rays targeted to make them look like photos of digital x-rays and x-ray films. We release this dataset as a resource for testing and improving the robustness of deep learning algorithms for automated chest x-ray interpretation on smartphone photos of chest x-rays.

Hao Sheng, Xiao Chen, Jingyi Su et al.

How can we effectively leverage the domain knowledge from remote sensing to better segment agriculture land cover from satellite images? In this paper, we propose a novel, model-agnostic, data-fusion approach for vegetation-related computer vision tasks. Motivated by the various Vegetation Indices (VIs), which are introduced by domain experts, we systematically reviewed the VIs that are widely used in remote sensing and their feasibility to be incorporated in deep neural networks. To fully leverage the Near-Infrared channel, the traditional Red-Green-Blue channels, and Vegetation Index or its variants, we propose a Generalized Vegetation Index (GVI), a lightweight module that can be easily plugged into many neural network architectures to serve as an additional information input. To smoothly train models with our GVI, we developed an Additive Group Normalization (AGN) module that does not require extra parameters of the prescribed neural networks. Our approach has improved the IoUs of vegetation-related classes by 0.9-1.3 percent and consistently improves the overall mIoU by 2 percent on our baseline.

Mang Tik Chiu, Xingqian Xu, Kai Wang et al.

The first Agriculture-Vision Challenge aims to encourage research in developing novel and effective algorithms for agricultural pattern recognition from aerial images, especially for the semantic segmentation task associated with our challenge dataset. Around 57 participating teams from various countries compete to achieve state-of-the-art in aerial agriculture semantic segmentation. The Agriculture-Vision Challenge Dataset was employed, which comprises of 21,061 aerial and multi-spectral farmland images. This paper provides a summary of notable methods and results in the challenge. Our submission server and leaderboard will continue to open for researchers that are interested in this challenge dataset and task; the link can be found here.

Akshay Smit, Saahil Jain, Pranav Rajpurkar et al.

The extraction of labels from radiology text reports enables large-scale training of medical imaging models. Existing approaches to report labeling typically rely either on sophisticated feature engineering based on medical domain knowledge or manual annotations by experts. In this work, we introduce a BERT-based approach to medical image report labeling that exploits both the scale of available rule-based systems and the quality of expert annotations. We demonstrate superior performance of a biomedically pretrained BERT model first trained on annotations of a rule-based labeler and then finetuned on a small set of expert annotations augmented with automated backtranslation. We find that our final model, CheXbert, is able to outperform the previous best rules-based labeler with statistical significance, setting a new SOTA for report labeling on one of the largest datasets of chest x-rays.

Pranav Rajpurkar, Anirudh Joshi, Anuj Pareek et al.

Although there have been several recent advances in the application of deep learning algorithms to chest x-ray interpretation, we identify three major challenges for the translation of chest x-ray algorithms to the clinical setting. We examine the performance of the top 10 performing models on the CheXpert challenge leaderboard on three tasks: (1) TB detection, (2) pathology detection on photos of chest x-rays, and (3) pathology detection on data from an external institution. First, we find that the top 10 chest x-ray models on the CheXpert competition achieve an average AUC of 0.851 on the task of detecting TB on two public TB datasets without fine-tuning or including the TB labels in training data. Second, we find that the average performance of the models on photos of x-rays (AUC = 0.916) is similar to their performance on the original chest x-ray images (AUC = 0.924). Third, we find that the models tested on an external dataset either perform comparably to or exceed the average performance of radiologists. We believe that our investigation will inform rapid translation of deep learning algorithms to safe and effective clinical decision support tools that can be validated prospectively with large impact studies and clinical trials.

Sharon Zhou, Jiequan Zhang, Hang Jiang et al.

Data augmentation has led to substantial improvements in the performance and generalization of deep models, and remain a highly adaptable method to evolving model architectures and varying amounts of data---in particular, extremely scarce amounts of available training data. In this paper, we present a novel method of applying Mobius transformations to augment input images during training. Mobius transformations are bijective conformal maps that generalize image translation to operate over complex inversion in pixel space. As a result, Mobius transformations can operate on the sample level and preserve data labels. We show that the inclusion of Mobius transformations during training enables improved generalization over prior sample-level data augmentation techniques such as cutout and standard crop-and-flip transformations, most notably in low data regimes.

David Rolnick, Priya L. Donti, Lynn H. Kaack et al.

Climate change is one of the greatest challenges facing humanity, and we, as machine learning experts, may wonder how we can help. Here we describe how machine learning can be a powerful tool in reducing greenhouse gas emissions and helping society adapt to a changing climate. From smart grids to disaster management, we identify high impact problems where existing gaps can be filled by machine learning, in collaboration with other fields. Our recommendations encompass exciting research questions as well as promising business opportunities. We call on the machine learning community to join the global effort against climate change.

Jeremy Irvin, Pranav Rajpurkar, Michael Ko et al.

Large, labeled datasets have driven deep learning methods to achieve expert-level performance on a variety of medical imaging tasks. We present CheXpert, a large dataset that contains 224,316 chest radiographs of 65,240 patients. We design a labeler to automatically detect the presence of 14 observations in radiology reports, capturing uncertainties inherent in radiograph interpretation. We investigate different approaches to using the uncertainty labels for training convolutional neural networks that output the probability of these observations given the available frontal and lateral radiographs. On a validation set of 200 chest radiographic studies which were manually annotated by 3 board-certified radiologists, we find that different uncertainty approaches are useful for different pathologies. We then evaluate our best model on a test set composed of 500 chest radiographic studies annotated by a consensus of 5 board-certified radiologists, and compare the performance of our model to that of 3 additional radiologists in the detection of 5 selected pathologies. On Cardiomegaly, Edema, and Pleural Effusion, the model ROC and PR curves lie above all 3 radiologist operating points. We release the dataset to the public as a standard benchmark to evaluate performance of chest radiograph interpretation models. The dataset is freely available at https://stanfordmlgroup.github.io/competitions/chexpert .

Anand Avati, Stephen Pfohl, Chris Lin et al.

Identifying patients who will be discharged within 24 hours can improve hospital resource management and quality of care. We studied this problem using eight years of Electronic Health Records (EHR) data from Stanford Hospital. We fit models to predict 24 hour discharge across the entire inpatient population. The best performing models achieved an area under the receiver-operator characteristic curve (AUROC) of 0.85 and an AUPRC of 0.53 on a held out test set. This model was also well calibrated. Finally, we analyzed the utility of this model in a decision theoretic framework to identify regions of ROC space in which using the model increases expected utility compared to the trivial always negative or always positive classifiers.

Anand Avati, Tony Duan, Sharon Zhou et al.

Probabilistic survival predictions from models trained with Maximum Likelihood Estimation (MLE) can have high, and sometimes unacceptably high variance. The field of meteorology, where the paradigm of maximizing sharpness subject to calibration is popular, has addressed this problem by using scoring rules beyond MLE, such as the Continuous Ranked Probability Score (CRPS). In this paper we present the \emph{Survival-CRPS}, a generalization of the CRPS to the survival prediction setting, with right-censored and interval-censored variants. We evaluate our ideas on the mortality prediction task using two different Electronic Health Record (EHR) data sets (STARR and MIMIC-III) covering millions of patients, with suitable deep neural network architectures: a Recurrent Neural Network (RNN) for STARR and a Fully Connected Network (FCN) for MIMIC-III. We compare results between the two scoring rules while keeping the network architecture and data fixed, and show that models trained with Survival-CRPS result in sharper predictive distributions compared to those trained by MLE, while still maintaining calibration.

Pranav Rajpurkar, Jeremy Irvin, Aarti Bagul et al.

We introduce MURA, a large dataset of musculoskeletal radiographs containing 40,561 images from 14,863 studies, where each study is manually labeled by radiologists as either normal or abnormal. To evaluate models robustly and to get an estimate of radiologist performance, we collect additional labels from six board-certified Stanford radiologists on the test set, consisting of 207 musculoskeletal studies. On this test set, the majority vote of a group of three radiologists serves as gold standard. We train a 169-layer DenseNet baseline model to detect and localize abnormalities. Our model achieves an AUROC of 0.929, with an operating point of 0.815 sensitivity and 0.887 specificity. We compare our model and radiologists on the Cohen's kappa statistic, which expresses the agreement of our model and of each radiologist with the gold standard. Model performance is comparable to the best radiologist performance in detecting abnormalities on finger and wrist studies. However, model performance is lower than best radiologist performance in detecting abnormalities on elbow, forearm, hand, humerus, and shoulder studies. We believe that the task is a good challenge for future research. To encourage advances, we have made our dataset freely available at https://stanfordmlgroup.github.io/competitions/mura .

Pranav Rajpurkar, Jeremy Irvin, Kaylie Zhu et al.

We develop an algorithm that can detect pneumonia from chest X-rays at a level exceeding practicing radiologists. Our algorithm, CheXNet, is a 121-layer convolutional neural network trained on ChestX-ray14, currently the largest publicly available chest X-ray dataset, containing over 100,000 frontal-view X-ray images with 14 diseases. Four practicing academic radiologists annotate a test set, on which we compare the performance of CheXNet to that of radiologists. We find that CheXNet exceeds average radiologist performance on the F1 metric. We extend CheXNet to detect all 14 diseases in ChestX-ray14 and achieve state of the art results on all 14 diseases.