Multidisciplinary tumor boards (MTBs) bring specialists together to identify therapies for complex cancer cases, but preparing for them is time-intensive. Clinicians must extract key details from extensive records and evaluate treatment options. While large language models (LLMs) show promise in medicine for basic tasks like summarizing notes, little is known about their role in high-stakes tasks like MTB preparation. We conducted a mixed-methods study with 16 oncologists using two AI systems to prepare patient cases for MTB: an off-the-shelf assistant (Copilot) and a task-specific multi-agent system (Healthcare Agent Orchestrator, HAO). We analyzed oncologist prompts, AI responses, and oncologists’ perception of AI. Participants showed greater willingness to adopt HAO but were often overconfident in AI summaries and skeptical of AI-recommended therapies. Trust calibration strategies, such as source links and agent-trajectories, failed to align trust with system capabilities. We conclude with how AI systems should be built to support clinicians in high-stakes tasks.

Dialogue systems are designed to offer human users social support or functional services through natural language interactions. Traditional conversation research has put significant emphasis on a system’s response-ability, including its capacity to understand dialogue context and generate appropriate responses. However, the key element of proactive behavior—a crucial aspect of intelligent conversations—is often overlooked in these studies. Proactivity empowers conversational agents to lead conversations towards achieving pre-defined targets or fulfilling specific goals on the system side. Proactive dialogue systems are equipped with advanced techniques to handle complex tasks, requiring strategic and motivational interactions, thus representing a significant step towards artificial general intelligence. Motivated by the necessity and challenges of building proactive dialogue systems, we provide a comprehensive review of various prominent problems and advanced designs for implementing proactivity into different types of dialogue systems, including open-domain dialogues, task-oriented dialogues, and information-seeking dialogues. We also discuss real-world challenges that require further research attention to meet application needs in the future, such as proactivity in dialogue systems that are based on large language models, proactivity in hybrid dialogues, evaluation protocols and ethical considerations for proactive dialogue systems. By providing a quick access and overall picture of the proactive dialogue systems domain, we aim to inspire new research directions and stimulate further advancements towards achieving the next level of conversational AI capabilities, paving the way for more dynamic and intelligent interactions within various application domains.

BACKGROUND: The use of real-world data (RWD) in artificial intelligence (AI) applications for healthcare offers unique opportunities but also poses complex challenges related to interpretability, transparency, safety, efficacy, bias, equity, privacy, ethics, accountability, and stakeholder engagement. METHODS: A multi-stakeholder expert panel comprising healthcare professionals, AI developers, policymakers, and other stakeholders was assembled. Their task was to identify critical issues and formulate consensus recommendations, focusing on the responsible use of RWD in healthcare AI. The panel's work involved an in-person conference and workshop and extensive deliberations over several months. RESULTS: The panel's findings revealed several critical challenges, including the necessity for data literacy and documentation, the identification and mitigation of bias, privacy and ethics considerations, and the absence of an accountability structure for stakeholder management. To address these, the panel proposed a series of recommendations, such as the adoption of metadata standards for RWD sources, the development of transparency frameworks and instructional labels likened to "nutrition labels" for AI applications, the provision of cross-disciplinary training materials, the implementation of bias detection and mitigation strategies, and the establishment of ongoing monitoring and update processes. CONCLUSION: Guidelines and resources focused on the responsible use of RWD in healthcare AI are essential for developing safe, effective, equitable, and trustworthy applications. The proposed recommendations provide a foundation for a comprehensive framework addressing the entire lifecycle of healthcare AI, emphasizing the importance of documentation, training, transparency, accountability, and multi-stakeholder engagement.

Yiqing Xie, Sheng Zhang, Hao Cheng, Pengfei Liu, Zelalem Gero, Cliff Wong, Tristan Naumann, Hoifung Poon, Carolyn Rose. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Pre-trained transformers are often fine-tuned to aid clinical decision-making using limited clinical notes. Model interpretability is crucial, especially in high-stakes domains like medicine, to establish trust and ensure safety, which requires human engagement. We introduce SUFO, a systematic framework that enhances interpretability of fine-tuned transformer feature spaces. SUFO utilizes a range of analytic and visualization techniques, including Supervised probing, Unsupervised similarity analysis, Feature dynamics, and Outlier analysis to address key questions about model trust and interpretability (e.g. model suitability for a task, feature space evolution during fine-tuning, and interpretation of fine-tuned features and failure modes). We conduct a case study investigating the impact of pre-training data where we focus on real-world pathology classification tasks, and validate our findings on MedNLI. We evaluate five 110M-sized pre-trained transformer models, categorized into general-domain (BERT, TNLR), mixed-domain (BioBERT, Clinical BioBERT), and domain-specific (PubMedBERT) groups. Our SUFO analyses reveal that: (1) while PubMedBERT, the domain-specific model, contains valuable information for fine-tuning, it can overfit to minority classes when class imbalances exist. In contrast, mixed-domain models exhibit greater resistance to overfitting, suggesting potential improvements in domain-specific model robustness; (2) in-domain pre-training accelerates feature disambiguation during fine-tuning; and (3) feature spaces undergo significant sparsification during this process, enabling clinicians to identify common outlier modes among fine-tuned models as demonstrated in this paper. These findings showcase the utility of SUFO in enhancing trust and safety when using transformers in medicine, and we believe SUFO can aid practitioners in evaluating fine-tuned language models (LMs) for other applications in medicine and in more critical domains.

BACKGROUND: Integrating artificial intelligence (AI) in healthcare settings has the potential to benefit clinical decision-making. Addressing challenges such as ensuring trustworthiness, mitigating bias, and maintaining safety is paramount. The lack of established methodologies for pre- and post-deployment evaluation of AI tools regarding crucial attributes such as transparency, performance monitoring, and adverse event reporting makes this situation challenging. OBJECTIVES: This paper aims to make practical suggestions for creating methods, rules, and guidelines to ensure that the development, testing, supervision, and use of AI in clinical decision support (CDS) systems are done well and safely for patients. MATERIALS AND METHODS: In May 2023, the Division of Clinical Informatics at Beth Israel Deaconess Medical Center and the American Medical Informatics Association co-sponsored a working group on AI in healthcare. In August 2023, there were 4 webinars on AI topics and a 2-day workshop in September 2023 for consensus-building. The event included over 200 industry stakeholders, including clinicians, software developers, academics, ethicists, attorneys, government policy experts, scientists, and patients. The goal was to identify challenges associated with the trusted use of AI-enabled CDS in medical practice. Key issues were identified, and solutions were proposed through qualitative analysis and a 4-month iterative consensus process. RESULTS: Our work culminated in several key recommendations: (1) building safe and trustworthy systems; (2) developing validation, verification, and certification processes for AI-CDS systems; (3) providing a means of safety monitoring and reporting at the national level; and (4) ensuring that appropriate documentation and end-user training are provided. DISCUSSION: AI-enabled Clinical Decision Support (AI-CDS) systems promise to revolutionize healthcare decision-making, necessitating a comprehensive framework for their development, implementation, and regulation that emphasizes trustworthiness, transparency, and safety. This framework encompasses various aspects including model training, explainability, validation, certification, monitoring, and continuous evaluation, while also addressing challenges such as data privacy, fairness, and the need for regulatory oversight to ensure responsible integration of AI into clinical workflow. CONCLUSIONS: Achieving responsible AI-CDS systems requires a collective effort from many healthcare stakeholders. This involves implementing robust safety, monitoring, and transparency measures while fostering innovation. Future steps include testing and piloting proposed trust mechanisms, such as safety reporting protocols, and establishing best practice guidelines.

one setup is more effective than another. In this work, we uncover the underlying characteristics of effective sets. For each training instance, we form a large, diverse pool of candidates and systematically vary the subsets used for calibration fine-tuning. Each selection strategy targets distinct aspects of the sets, such as lexical diversity or the size of the gap between positive and negatives. On three diverse scientific long-form summarization datasets (spanning biomedical, clinical, and chemical domains), we find, among others, that faithfulness calibration is optimal when the negative sets are extractive and more likely to be generated, whereas for relevance calibration, the metric margin between candidates should be maximized and surprise-the disagreement between model and metric defined candidate rankings-minimized. Code to create, select, and optimize calibration sets is available at https://github.com/griff4692/calibrating-summaries.

Relation extraction (RE), which has relied on structurally annotated corpora for model training, has been particularly challenging in lowresource scenarios and domains.Recent literature has tackled low-resource RE by selfsupervised learning, where the solution involves pretraining the entity pair embedding by RE-based objective and finetuning on labeled data by classification-based objective.However, a critical challenge to this approach is the gap in objectives, which prevents the RE model from fully utilizing the knowledge in pretrained representations.In this paper, we aim at bridging the gap and propose to pretrain and finetune the RE model using consistent objectives of contrastive learning.Since in this kind of representation learning paradigm, one relation may easily form multiple clusters in the representation space, we further propose a multi-center contrastive loss that allows one relation to form multiple clusters to better align with pretraining.Experiments on two document-level RE datasets, BioRED and Re-DocRED, demonstrate the effectiveness of our method.Particularly, when using 1% end-task training data, our method outperforms PLMbased RE classifier by 10.5% and 6.1% on the two datasets, respectively.

Medicine today is imprecise. Among the top 20 drugs in the U.S., up to 80% of patients are non-responders. The goal of precision health is to provide the right intervention for the right people at the right time. The key to realize this dream is to develop a data-driven, learning system that can instantly incorporate new health information to optimize care delivery and accelerate biomedical discovery. In reality, however, the health ecosystem is mired in overwhelming unstructured data and excruciating manual processing. For example, in cancer, standard of care often fails, and clinical trials are the last hope. Yet less than 3% of patients could find a matching trial, whereas 40% of trial failures simply stem from insufficient recruitment. Discovery is painfully slow as a new drug may take billions of dollars and over a decade to develop.

Conversational generative AI has demonstrated remarkable promise for empowering biomedical practitioners, but current investigations focus on unimodal text. Multimodal conversational AI has seen rapid progress by leveraging billions of image-text pairs from the public web, but such general-domain vision-language models still lack sophistication in understanding and conversing about biomedical images. In this paper, we propose a cost-efficient approach for training a vision-language conversational assistant that can answer open-ended research questions of biomedical images. The key idea is to leverage a large-scale, broad-coverage biomedical figure-caption dataset extracted from PubMed Central, use GPT-4 to self-instruct open-ended instruction-following data from the captions, and then fine-tune a large general-domain vision-language model using a novel curriculum learning method. Specifically, the model first learns to align biomedical vocabulary using the figure-caption pairs as is, then learns to master open-ended conversational semantics using GPT-4 generated instruction-following data, broadly mimicking how a layperson gradually acquires biomedical knowledge. This enables us to train a Large Language and Vision Assistant for BioMedicine (LLaVA-Med) in less than 15 hours (with eight A100s). LLaVA-Med exhibits excellent multimodal conversational capability and can follow open-ended instruction to assist with inquiries about a biomedical image. On three standard biomedical visual question answering datasets, LLaVA-Med outperforms previous supervised state-of-the-art on certain metrics. To facilitate biomedical multimodal research, we will release our instruction-following data and the LLaVA-Med model.

Abstract Label scarcity is a bottleneck for improving task performance in specialized domains. We propose a novel compositional transfer learning framework (DoT51) for zero-shot domain transfer. Without access to in-domain labels, DoT5 jointly learns domain knowledge (from masked language modelling of unlabelled in-domain free text) and task knowledge (from task training on more readily available general-domain data) in a multi-task manner. To improve the transferability of task training, we design a strategy named NLGU: We simultaneously train natural language generation (NLG) for in-domain label-to-data generation, which enables data augmentation for self-finetuning and natural language understanding (NLU) for label prediction. We evaluate DoT5 on the biomedical domain and the resource-lean subdomain of radiology, focusing on natural language inference, text summarization, and embedding learning. DoT5 demonstrates the effectiveness of compositional transfer learning through multi-task learning. In particular, DoT5 outperforms the current state-of-the-art in zero-shot transfer by over 7 absolute points in accuracy on RadNLI. We validate DoT5 with ablations and a case study demonstrating its ability to solve challenging NLI examples requiring in-domain expertise.

Pretraining large neural language models, such as BERT, has led to impressive gains on many natural language processing (NLP) tasks. However, most pretraining efforts focus on general domain corpora, such as newswire and Web. A prevailing assumption is that even domain-specific pretraining can benefit by starting from general-domain language models. In this article, we challenge this assumption by showing that for domains with abundant unlabeled text, such as biomedicine, pretraining language models from scratch results in substantial gains over continual pretraining of general-domain language models. To facilitate this investigation, we compile a comprehensive biomedical NLP benchmark from publicly available datasets. Our experiments show that domain-specific pretraining serves as a solid foundation for a wide range of biomedical NLP tasks, leading to new state-of-the-art results across the board. Further, in conducting a thorough evaluation of modeling choices, both for pretraining and task-specific fine-tuning, we discover that some common practices are unnecessary with BERT models, such as using complex tagging schemes in named entity recognition. To help accelerate research in biomedical NLP, we have released our state-of-the-art pretrained and task-specific models for the community, and created a leaderboard featuring our BLURB benchmark (short for Biomedical Language Understanding & Reasoning Benchmark) at https://aka.ms/BLURB .

Extracting relations across large text spans has been relatively underexplored in NLP, but it is particularly important for high-value domains such as biomedicine, where obtaining high recall of the latest findings is crucial for practical applications.Compared to conventional information extraction confined to short text spans, document-level relation extraction faces additional challenges in both inference and learning.Given longer text spans, state-of-the-art neural architectures are less effective and taskspecific self-supervision such as distant supervision becomes very noisy.In this paper, we propose decomposing document-level relation extraction into relation detection and argument resolution, taking inspiration from Davidsonian semantics.This enables us to incorporate explicit discourse modeling and leverage modular self-supervision for each sub-problem, which is less noise-prone and can be further refined end-to-end via variational EM.We conduct a thorough evaluation in biomedical machine reading for precision oncology, where cross-paragraph relation mentions are prevalent.Our method outperforms prior state of the art, such as multi-scale learning and graph neural networks, by over 20 absolute F1 points.The gain is particularly pronounced among the most challenging relation instances whose arguments never co-occur in a paragraph.

Information overload is a prevalent challenge in many high-value domains. A prominent case in point is the explosion of the biomedical literature on COVID-19, which swelled to hundreds of thousands of papers in a matter of months. In general, biomedical literature expands by two papers every minute, totalling over a million new papers every year. Search in the biomedical realm, and many other vertical domains is challenging due to the scarcity of direct supervision from click logs. Self-supervised learning has emerged as a promising direction to overcome the annotation bottleneck. We propose a general approach for vertical search based on domain-specific pretraining and present a case study for the biomedical domain. Despite being substantially simpler and not using any relevance labels for training or development, our method performs comparably or better than the best systems in the official TREC-COVID evaluation, a COVID-related biomedical search competition. Using distributed computing in modern cloud infrastructure, our system can scale to tens of millions of articles on PubMed and has been deployed as Microsoft Biomedical Search, a new search experience for biomedical literature: https://aka.ms/biomedsearch.

This article reviews recent advances in applying natural language processing (NLP) to Electronic Health Records (EHRs) for computational phenotyping. NLP-based computational phenotyping has numerous applications including diagnosis categorization, novel phenotype discovery, clinical trial screening, pharmacogenomics, drug-drug interaction (DDI), and adverse drug event (ADE) detection, as well as genome-wide and phenome-wide association studies. Significant progress has been made in algorithm development and resource construction for computational phenotyping. Among the surveyed methods, well-designed keyword search and rule-based systems often achieve good performance. However, the construction of keyword and rule lists requires significant manual effort, which is difficult to scale. Supervised machine learning models have been favored because they are capable of acquiring both classification patterns and structures from data. Recently, deep learning and unsupervised learning have received growing attention, with the former favored for its performance and the latter for its ability to find novel phenotypes. Integrating heterogeneous data sources have become increasingly important and have shown promise in improving model performance. Often, better performance is achieved by combining multiple modalities of information. Despite these many advances, challenges and opportunities remain for NLP-based computational phenotyping, including better model interpretability and generalizability, and proper characterization of feature relations in clinical narratives.

The biomedical field offers many learning tasks that share unique challenges: large amounts of unpaired data, and a high cost to generate labels. In this work, we develop a method to address these issues with semi-supervised learning in regression tasks (e.g., translation from source to target). Our model uses adversarial signals to learn from unpaired datapoints, and imposes a cycle-loss reconstruction error penalty to regularize mappings in either direction against one another. We first evaluate our method on synthetic experiments, demonstrating two primary advantages of the system: 1) distribution matching via the adversarial loss and 2) regularization towards invertible mappings via the cycle loss. We then show a regularization effect and improved performance when paired data is supplemented by additional unpaired data on two real biomedical regression tasks: estimating the physiological effect of medical treatments, and extrapolating gene expression (transcriptomics) signals. Our proposed technique is a promising initial step towards more robust use of adversarial signals in semi-supervised regression, and could be useful for other tasks (e.g., causal inference or modality translation) in the biomedical field.

Existing machine learning methods typically assume consistency in how semantically equivalent information is encoded. However, the way information is recorded in databases differs across institutions and over time, often rendering potentially useful data obsolescent. To address this problem, we map database-specific representations of information to a shared set of semantic concepts, thus allowing models to be built from or transition across different databases. We demonstrate our method on machine learning models developed in a healthcare setting. In particular, we evaluate our method using two different intensive care unit (ICU) databases and on two clinically relevant tasks, in-hospital mortality and prolonged length of stay. For both outcomes, a feature representation mapping EHR-specific events to a shared set of clinical concepts yields better results than using EHR-specific events alone.

The ability to determine patient acuity (or severity of illness) has immediate practical use for clinicians. We evaluate the use of multivariate timeseries modeling with the multi-task Gaussian process (GP) models using noisy, incomplete, sparse, heterogeneous and unevenly-sampled clinical data, including both physiological signals and clinical notes. The learned multi-task GP (MTGP) hyperparameters are then used to assess and forecast patient acuity. Experiments were conducted with two real clinical data sets acquired from ICU patients: firstly, estimating cerebrovascular pressure reactivity, an important indicator of secondary damage for traumatic brain injury patients, by learning the interactions between intracranial pressure and mean arterial blood pressure signals, and secondly, mortality prediction using clinical progress notes. In both cases, MTGPs provided improved results: an MTGP model provided better results than single-task GP models for signal interpolation and forecasting (0.91 vs 0.69 RMSE), and the use of MTGP hyperparameters obtained improved results when used as additional classification features (0.812 vs 0.788 AUC).

Accurate knowledge of a patient's disease state and trajectory is critical in a clinical setting. Modern electronic healthcare records contain an increasingly large amount of data, and the ability to automatically identify the factors that influence patient outcomes stand to greatly improve the efficiency and quality of care. We examined the use of latent variable models (viz. Latent Dirichlet Allocation) to decompose free-text hospital notes into meaningful features, and the predictive power of these features for patient mortality. We considered three prediction regimes: (1) baseline prediction, (2) dynamic (time-varying) outcome prediction, and (3) retrospective outcome prediction. In each, our prediction task differs from the familiar time-varying situation whereby data accumulates; since fewer patients have long ICU stays, as we move forward in time fewer patients are available and the prediction task becomes increasingly difficult. We found that latent topic-derived features were effective in determining patient mortality under three timelines: inhospital, 30 day post-discharge, and 1 year post-discharge mortality. Our results demonstrated that the latent topic features important in predicting hospital mortality are very different from those that are important in post-discharge mortality. In general, latent topic features were more predictive than structured features, and a combination of the two performed best. The time-varying models that combined latent topic features and baseline features had AUCs that reached 0.85, 0.80, and 0.77 for in-hospital, 30 day post-discharge and 1 year post-discharge mortality respectively. Our results agreed with other work suggesting that the first 24 hours of patient information are often the most predictive of hospital mortality. Retrospective models that used a combination of latent topic features and structured features achieved AUCs of 0.96, 0.82, and 0.81 for in-hospital, 30 day, and 1-year mortality prediction. Our work focuses on the dynamic (time-varying) setting because models from this regime could facilitate an on-going severity stratification system that helps direct care-staff resources and inform treatment strategies.