Medical coding translates clinical documentation into standardized codes for billing, research, and public health, but manual coding is time-consuming and error-prone. Existing automation efforts rely on small datasets that poorly represent real-world patient heterogeneity. We trained a language model on 5.8 million electronic health records from 1.8 million patients across nearly all specialties in Eastern Denmark (2006–2016) to predict ICD-10 codes from clinical notes, medications, and laboratory results. Evaluated on 270,000 held-out patients, the model achieved a micro F1 of 71.8

Modern neural networks have greatly improved performance across speech recognition benchmarks. However, gains are often driven by frequent words with limited semantic weight, which can obscure meaningful differences in word error rate, the primary evaluation metric. Errors in rare terms, named entities, and domain-specific vocabulary are more consequential, but remain hidden by aggregate metrics. This highlights the need for finer-grained error analysis, which depends on accurate alignment between reference and model transcripts. However, conventional alignment methods are not designed for such precision. We propose a novel alignment algorithm that couples dynamic programming with beam search scoring. Compared to traditional text alignment methods, our approach provides more accurate alignment of individual errors, enabling reliable error analysis. The algorithm is made available via PyPI.

After decades of use in dictation and, more recently, ambient documentation, speech is emerging as a primary modality for interacting with technology and AI in healthcare. Yet medical speech recognition remains difficult: systems must capture specialized terminology, resolve contextual ambiguity, and render measurements, abbreviations, and clinical shorthand precisely. Existing solutions are typically optimized either for general-purpose transcription or narrow dictation workflows, limiting their reliability in safety-critical settings and their usefulness for broader clinical workflows. We introduce Symphony for Speech-to-Text, a medical-grade speech recognition system for real-time streaming and batch file-based clinical use. Symphony decomposes the transcription process into specialized components for recognition, formatting, and contextual correction to optimize medical term recall while producing clinically structured text in real time and adapting across use cases. Evaluations on public benchmark and medical speech datasets show that Symphony substantially outperforms state-of-the-art systems in clinical settings while matching or exceeding them in general-domain settings, suggesting robust generalization rather than overfitting. We release a clinical benchmark dataset to support reliable validation and further progress in medical speech recognition. Symphony is available through a production-grade API for live dictation, conversational transcription, and batch audio file processing.

Medical coding translates free-text clinical documentation into standardized codes drawn from classification systems that contain tens of thousands of entries and are updated annually. It is central to billing, clinical research, and quality reporting, yet remains largely manual, slow, and error-prone. Existing automated approaches learn to predict a fixed set of codes from labeled data, thereby preventing adaptation to new codes or different coding systems without retraining on different data. They also provide no explanation for their predictions, limiting trust in safety-critical settings. We introduce Symphony for Medical Coding, a system that approaches the task the way expert human coders do: by reasoning over the clinical narrative with direct access to the coding guidelines. This design allows Symphony to operate across any coding system and to provide span-level evidence linking each predicted code to the text that supports it. We evaluate on two public benchmarks and three real-world datasets spanning inpatient, outpatient, emergency, and subspecialty settings across the United States and the United Kingdom. Symphony achieves state-of-the-art results across all settings, establishing itself as a flexible, deployment-ready foundation for automated clinical coding.

In medical coding, experts map unstructured clinical notes to alphanumeric codes for diagnoses and procedures. We introduce Code Like Humans: a new agentic framework for medical coding with large language models. It implements official coding guidelines for human experts, and it is the first solution that can support the full ICD-10 coding system (+70K labels). It achieves the best performance to date on rare diagnosis codes (fine-tuned discriminative classifiers retain an advantage for high-frequency codes, to which they are limited). Towards future work, we also contribute an analysis of system performance and identify its `blind spots' (codes that are systematically undercoded).

Deep neural network predictions are notoriously difficult to interpret. Feature attribution methods aim to explain these predictions by identifying the contribution of each input feature. Faithfulness, often evaluated using the area over the perturbation curve (AOPC), reflects feature attributions’ accuracy in describing the internal mechanisms of deep neural networks. However, many studies rely on AOPC to compare faithfulness across different models, which we show can lead to false conclusions about models’ faithfulness. Specifically, we find that AOPC is sensitive to variations in the model, resulting in unreliable cross-model comparisons. Moreover, AOPC scores are difficult to interpret in isolation without knowing the model-specific lower and upper limits. To address these issues, we propose a normalization approach, Normalized AOPC (NAOPC), enabling consistent cross-model evaluations and more meaningful interpretation of individual scores. Our experiments demonstrate that this normalization can radically change AOPC results, questioning the conclusions of earlier studies and offering a more robust framework for assessing feature attribution faithfulness. Our code is available at https://github.com/JoakimEdin/naopc.

Background: Chronic obstructive pulmonary disease (COPD) is projected to be the third-leading cause of death by 2030. Traditional spirometry for the monitoring of the forced expiratory volume in one second (FEV1) can provoke discomfort and anxiety. This study aimed to validate AI models using daily audio recordings as an alternative for FEV1 estimation in home settings. Methods: Twenty-three participants with moderate to severe COPD recorded daily audio readings of standardized texts and measured their FEV1 using spirometry over nine months. Participants also recorded biomarkers (heart rate, temperature, oxygen saturation) via tablet application. Various machine learning models were trained using acoustic features extracted from 2053 recordings, with K-nearest neighbor, random forest, XGBoost, and linear models evaluated using 10-fold cross-validation. Results: The K-nearest neighbors model achieved a root mean square error of 174 mL/s on the validation data. The limit of agreement (LoA) ranged from −333.21 to 347.26 mL/s. Despite an error range of −1252 to 1435 mL/s, most predictions fell within the LoA, indicating good performance in estimating the FEV1. Conclusions: The predictive model showed promising results, with a narrower LoA compared to traditional unsupervised spirometry methods. The AI models effectively used audio to predict the FEV1, suggesting a viable non-invasive approach for COPD monitoring that could enhance patient comfort and accessibility in home settings.

Circuit localization methods aim to identify the subset of model components responsible for specific behaviors in large language models, enabling detailed mechanistic analysis. Most existing methods assume components act independently and estimate importance by perturbing each component in isolation. However, components in neural networks interact, and ignoring these interactions leads to systematic misestimation of component importance. We find that one particularly problematic interaction is attention self-repair, in which softmax redistribution causes gradients for influential attention scores to vanish as other positions with similar values compensate. We introduce Gradient Interaction Modifications (GIM), a technique that explicitly accounts for feature interactions during backpropagation. GIM achieves state-of-the-art performance on the circuit localization track of the Mechanistic Interpretability Benchmark and outperforms existing gradient-based methods on feature attribution across diverse tasks. By accounting for interaction effects and explaining why prior methods underestimate component importance, GIM enables more faithful mechanistic analysis of large language models. GIM is available as a Python package at https://github.com/corticph/gim.

Electronic healthcare records are vital for patient safety as they document conditions, plans, and procedures in both free text and medical codes. Language models have significantly enhanced the processing of such records, streamlining workflows and reducing manual data entry, thereby saving healthcare providers significant resources. However, the black-box nature of these models often leaves healthcare professionals hesitant to trust them. State-of-the-art explainability methods increase model transparency but rely on human-annotated evidence spans, which are costly. In this study, we propose an approach to produce plausible and faithful explanations without needing such annotations. We demonstrate on the automated medical coding task that adversarial robustness training improves explanation plausibility and introduce AttInGrad, a new explanation method superior to previous ones. By combining both contributions in a fully unsupervised setup, we produce explanations of comparable quality, or better, to that of a supervised approach. We release our code and model weights.

Medical coding is the task of assigning medical codes to clinical free-text documentation. Healthcare professionals manually assign such codes to track patient diagnoses and treatments. Automated medical coding can considerably alleviate this administrative burden. In this paper, we reproduce, compare, and analyze state-of-the-art automated medical coding machine learning models. We show that several models underperform due to weak configurations, poorly sampled train-test splits, and insufficient evaluation. In previous work, the macro F1 score has been calculated sub-optimally, and our correction doubles it. We contribute a revised model comparison using stratified sampling and identical experimental setups, including hyperparameters and decision boundary tuning. We analyze prediction errors to validate and falsify assumptions of previous works. The analysis confirms that all models struggle with rare codes, while long documents only have a negligible impact. Finally, we present the first comprehensive results on the newly released MIMIC-IV dataset using the reproduced models. We release our code, model parameters, and new MIMIC-III and MIMIC-IV training and evaluation pipelines to accommodate fair future comparisons.

Advanced stroke treatment is time-dependent and, therefore, relies on recognition by call-takers at prehospital telehealth services to ensure fast hospitalisation. This study aims to develop and assess the potential of machine learning in improving prehospital stroke recognition during medical helpline calls. We used calls from 1 January 2015 to 31 December 2020 in Copenhagen to develop a machine learning-based classification pipeline. Calls from 2021 are used for testing. Calls are first transcribed using an automatic speech recognition model and then categorised as stroke or non-stroke using a text classification model. Call-takers achieve a sensitivity of 52.7% (95% confidence interval 49.2–56.4%) with a positive predictive value (PPV) of 17.1% (15.5–18.6%). The machine learning framework performs significantly better ( p < 0.0001) with a sensitivity of 63.0% (62.0–64.1%) and a PPV of 24.9% (24.3–25.5%). Thus, a machine learning framework for recognising stroke in prehospital medical helpline calls may become a supportive tool for call-takers, aiding in early and accurate stroke recognition.

Although supervised deep learning has revolutionized speech and audio processing, it has necessitated the building of specialist models for individual tasks and application scenarios. It is likewise difficult to apply this to dialects and languages for which only limited labeled data is available. Self-supervised representation learning methods promise a single universal model that would benefit a wide variety of tasks and domains. Such methods have shown success in natural language processing and computer vision domains, achieving new levels of performance while reducing the number of labels required for many downstream scenarios. Speech representation learning is experiencing similar progress in three main categories: generative, contrastive, and predictive methods. Other approaches rely on multi-modal data for pre-training, mixing text or visual data streams with speech. Although self-supervised speech representation is still a nascent research area, it is closely related to acoustic word embedding and learning with zero lexical resources, both of which have seen active research for many years. This review presents approaches for self-supervised speech representation learning and their connection to other research areas. Since many current methods focus solely on automatic speech recognition as a downstream task, we review recent efforts on benchmarking learned representations to extend the application beyond speech recognition.

Introduction:Early recognition of out-of-hospital cardiac arrest (OHCA) by ambulance service call centre operators is important so that cardiopulmonary resuscitation can be delivered immediately, but around 25% of OHCAs are not picked up by call centre operators. An artificial intelligence (AI) system has been developed to support call centre operators in the detection of OHCA. The study aims to (1) explore ambulance service stakeholder perceptions on the safety of OHCA AI decision support in call centres, and (2) develop a clinical safety case for the OHCA AI decision-support system.Methods and analysis:The study will be undertaken within the Welsh Ambulance Service. The study is part research and part service evaluation. The research utilises a qualitative study design based on thematic analysis of interview data. The service evaluation consists of the development of a clinical safety case based on document analysis, analysis of the AI model and its development process and informal interviews with the technology developer.Conclusions:AI presents many opportunities for ambulance services, but safety assurance requirements need to be understood. The ASSIST project will continue to explore and build the body of knowledge in this area.

Unsupervised representation learning for speech processing has matured greatly in the last few years. Work in computer vision and natural language processing has paved the way, but speech data offers unique challenges. As a result, methods from other domains rarely translate directly. We review the development of unsupervised representation learning for speech over the last decade. We identify two primary model categories: self-supervised methods and probabilistic latent variable models. We describe the models and develop a comprehensive taxonomy. Finally, we discuss and compare models from the two categories.

Stochastic latent variable models (LVMs) achieve state-of-the-art performance on natural image generation but are still inferior to deterministic models on speech. In this paper, we develop a speech benchmark of popular temporal LVMs and compare them against state-of-the-art deterministic models. We report the likelihood, which is a much used metric in the image domain, but rarely, or incomparably, reported for speech models. To assess the quality of the learned representations, we also compare their usefulness for phoneme recognition. Finally, we adapt the Clockwork VAE, a state-of-the-art temporal LVM for video generation, to the speech domain. Despite being autoregressive only in latent space, we find that the Clockwork VAE can outperform previous LVMs and reduce the gap to deterministic models by using a hierarchy of latent variables.

While stochastic latent variable models (LVMs) now achieve state-of-the-art performance on natural image generation, they are still inferior to deterministic models on speech. On natural images, these models have been parameterised with very deep hierarchies of latent variables, but research shows that these model constructs are not directly applicable to sequence data. In this paper, we benchmark popular temporal LVMs against state-of-the-art deterministic models on speech. We report the likelihood, which is a much used metric in the image domain but rarely, and often incomparably, reported for speech models. This is prerequisite work needed for the research community to improve LVMs on speech. We adapt Clockwork VAE, a state-of-the-art temporal LVM for video generation, to the speech domain, similar to how WaveNet adapted PixelCNN from images to speech. Despite being autoregressive only in latent space, we find that the Clockwork VAE outperforms previous LVMs and reduces the gap to deterministic models by using a hierarchy of latent variables.

We present simple methods for out-of-distribution detection using a trained generative model. These techniques, based on classical statistical tests, are model-agnostic in the sense that they can be applied to any differentiable generative model. The idea is to combine a classical parametric test (Rao's score test) with the recently introduced typicality test. These two test statistics are both theoretically well-founded and exploit different sources of information based on the likelihood for the typicality test and its gradient for the score test. We show that combining them using Fisher's method overall leads to a more accurate out-of-distribution test. We also discuss the benefits of casting out-of-distribution detection as a statistical testing problem, noting in particular that false positive rate control can be valuable for practical out-of-distribution detection. Despite their simplicity and generality, these methods can be competitive with model-specific out-of-distribution detection algorithms without any assumptions on the out-distribution.

Recent advances in self-supervised learning through contrastive training have shown that it is possible to learn a competitive speech recognition system with as little as 10 minutes of labeled data. However, these systems are computationally expensive since they require pre-training followed by fine-tuning in a large parameter space. We explore the performance of such systems without fine-tuning by training a state-of-the-art speech recognizer on the fixed representations from the computationally demanding wav2vec 2.0 framework. We find performance to decrease without fine-tuning and, in the extreme low-resource setting, wav2vec 2.0 is inferior to its predecessor. In addition, we find that wav2vec 2.0 representations live in a low dimensional subspace and that decorrelating the features of the representations can stabilize training of the automatic speech recognizer. Finally, we propose a bidirectional extension to the original wav2vec framework that consistently improves performance.

Spoken language understanding (SLU) tasks are usually solved by first transcribing an utterance with automatic speech recognition (ASR) and then feeding the output to a text-based model. Recent advances in self-supervised representation learning for speech data have focused on improving the ASR component. We investigate whether representation learning for speech has matured enough to replace ASR in SLU. We compare learned speech features from wav2vec 2.0, state-of-the-art ASR transcripts, and the ground truth text as input for a novel speech-based named entity recognition task, a cardiac arrest detection task on real-world emergency calls and two existing SLU benchmarks. We show that learned speech features are superior to ASR transcripts on three classification tasks. For machine translation, ASR transcripts are still the better choice. We highlight the intrinsic robustness of wav2vec 2.0 representations to out-of-vocabulary words as key to better performance.