Paul Schmiedmayer

Fan Wu, Tsai-Ning Wang, Nicolas Zumarraga et al. · eth-zurich, harvard

Auscultation is a vital diagnostic tool, yet its utility is often limited by subjective interpretation. While general-purpose Audio-Language Models (ALMs) excel in general domains, they struggle with the nuances of physiological signals. We propose a framework that aligns multi-site auscultation recordings directly with a frozen Large Language Model (LLM) embedding space via gated cross-attention. By leveraging the LLM's latent world knowledge, our approach moves beyond isolated classification toward holistic, patient-level assessment. On the CaReSound benchmark, our model achieves a state-of-the-art 0.865 F1-macro and 0.952 BERTScore. We demonstrate that lightweight, domain-specific encoders rival large-scale ALMs and that multi-site aggregation provides spatial redundancy that mitigates temporal truncation. This alignment of medical acoustics with text foundations offers a scalable path for bridging signal processing and clinical assessment.

Philipp Zagar, Vishnu Ravi, Lauren Aalami et al.

The ability of large language models (LLMs) to transform, interpret, and comprehend vast quantities of heterogeneous data presents a significant opportunity to enhance data-driven care delivery. However, the sensitive nature of protected health information (PHI) raises valid concerns about data privacy and trust in remote LLM platforms. In addition, the cost associated with cloud-based artificial intelligence (AI) services continues to impede widespread adoption. To address these challenges, we propose a shift in the LLM execution environment from opaque, centralized cloud providers to a decentralized and dynamic fog computing architecture. By executing open-weight LLMs in more trusted environments, such as the user's edge device or a fog layer within a local network, we aim to mitigate the privacy, trust, and financial challenges associated with cloud-based LLMs. We further present SpeziLLM, an open-source framework designed to facilitate rapid and seamless leveraging of different LLM execution layers and lowering barriers to LLM integration in digital health applications. We demonstrate SpeziLLM's broad applicability across six digital health applications, showcasing its versatility in various healthcare settings.

Patrick Langer, Thomas Kaar, Max Rosenblattl et al.

LLMs have emerged as powerful tools for interpreting multimodal data. In medicine, they hold particular promise for synthesizing large volumes of clinical information into actionable insights and digital health applications. Yet, a major limitation remains their inability to handle time series. To overcome this gap, we present OpenTSLM, a family of Time Series Language Models (TSLMs) created by integrating time series as a native modality to pretrained LLMs, enabling reasoning over multiple time series of any length. We investigate two architectures for OpenTSLM. The first, OpenTSLM-SoftPrompt, models time series implicitly by concatenating learnable time series tokens with text tokens via soft prompting. Although parameter-efficient, we hypothesize that explicit time series modeling scales better and outperforms implicit approaches. We thus introduce OpenTSLM-Flamingo, which integrates time series with text via cross-attention. We benchmark both variants against baselines that treat time series as text tokens or plots, across a suite of text-time-series Chain-of-Thought (CoT) reasoning tasks. We introduce three datasets: HAR-CoT, Sleep-CoT, and ECG-QA-CoT. Across all, OpenTSLM models outperform baselines, reaching 69.9 F1 in sleep staging and 65.4 in HAR, compared to 9.05 and 52.2 for finetuned text-only models. Notably, even 1B-parameter OpenTSLM models surpass GPT-4o (15.47 and 2.95). OpenTSLM-Flamingo matches OpenTSLM-SoftPrompt in performance and outperforms on longer sequences, while maintaining stable memory requirements. By contrast, SoftPrompt grows exponentially in memory with sequence length, requiring around 110 GB compared to 40 GB VRAM when training on ECG-QA with LLaMA-3B. Expert reviews by clinicians find strong reasoning capabilities exhibited by OpenTSLMs on ECG-QA. To facilitate further research, we provide all code, datasets, and models open-source.

Paul Schmiedmayer, Adrit Rao, Philipp Zagar et al.

Objective: To enhance health literacy and accessibility of health information for a diverse patient population by developing a patient-centered artificial intelligence (AI) solution using large language models (LLMs) and Fast Healthcare Interoperability Resources (FHIR) application programming interfaces (APIs). Materials and Methods: The research involved developing LLM on FHIR, an open-source mobile application allowing users to interact with their health records using LLMs. The app is built on Stanford's Spezi ecosystem and uses OpenAI's GPT-4. A pilot study was conducted with the SyntheticMass patient dataset and evaluated by medical experts to assess the app's effectiveness in increasing health literacy. The evaluation focused on the accuracy, relevance, and understandability of the LLM's responses to common patient questions. Results: LLM on FHIR demonstrated varying but generally high degrees of accuracy and relevance in providing understandable health information to patients. The app effectively translated medical data into patient-friendly language and was able to adapt its responses to different patient profiles. However, challenges included variability in LLM responses and the need for precise filtering of health data. Discussion and Conclusion: LLMs offer significant potential in improving health literacy and making health records more accessible. LLM on FHIR, as a pioneering application in this field, demonstrates the feasibility and challenges of integrating LLMs into patient care. While promising, the implementation and pilot also highlight risks such as inconsistent responses and the importance of replicable output. Future directions include better resource identification mechanisms and executing LLMs on-device to enhance privacy and reduce costs.

Leon Nissen, Philipp Zagar, Vishnu Ravi et al.

The deployment of Large Language Models (LLM) on mobile devices offers significant potential for medical applications, enhancing privacy, security, and cost-efficiency by eliminating reliance on cloud-based services and keeping sensitive health data local. However, the performance and accuracy of on-device LLMs in real-world medical contexts remain underexplored. In this study, we benchmark publicly available on-device LLMs using the AMEGA dataset, evaluating accuracy, computational efficiency, and thermal limitation across various mobile devices. Our results indicate that compact general-purpose models like Phi-3 Mini achieve a strong balance between speed and accuracy, while medically fine-tuned models such as Med42 and Aloe attain the highest accuracy. Notably, deploying LLMs on older devices remains feasible, with memory constraints posing a greater challenge than raw processing power. Our study underscores the potential of on-device LLMs for healthcare while emphasizing the need for more efficient inference and models tailored to real-world clinical reasoning.

Nicolas Zumarraga, Thomas Kaar, Ning Wang et al.

Time Series Language Models (TSLMs) are emerging as unified models for reasoning over continuous signals in natural language. However, long-context retrieval remains a major limitation: existing models are typically trained and evaluated on short sequences, while real-world time-series sensor streams can span millions of datapoints. This mismatch requires precise temporal localization under strict computational constraints, a regime that is not captured by current benchmarks. We introduce TS-Haystack, a long-context temporal retrieval benchmark comprising ten task types across four categories: direct retrieval, temporal reasoning, multi-step reasoning and contextual anomaly. The benchmark uses controlled needle insertion by embedding short activity bouts into longer longitudinal accelerometer recordings, enabling systematic evaluation across context lengths ranging from seconds to 2 hours per sample. We hypothesize that existing TSLM time series encoders overlook temporal granularity as context length increases, creating a task-dependent effect: compression aids classification but impairs retrieval of localized events. Across multiple model and encoding strategies, we observe a consistent divergence between classification and retrieval behavior. Learned latent compression preserves or improves classification accuracy at compression ratios up to 176$\times$, but retrieval performance degrades with context length, incurring in the loss of temporally localized information. These results highlight the importance of architectural designs that decouple sequence length from computational complexity while preserving temporal fidelity.