arXiv subjects

Subhash Nerella

Publications and source records attributed to Subhash Nerella.

17 recordsLinked to original sources

arXiv2026

Teaching LLMs How ICU Physicians Approach Clinical Reasoning Through OMOP-Aligned Retrieval Improves Reasoning Across Clinical Domains

Clinical decision-making relies on identifying relevant patient information to guide diagnosis and treatment, a challenge that is especially difficult in the data-dense and rapidly changing intensive care unit (ICU). Large language models (LLMs) could support this task. However, existing applications and datasets mostly emphasize surface-level retrieval or factual recall rather than the inductive and deductive reasoning clinicians practice to select and reason over decision-relevant evidence. We hypothesized that training LLMs on expert ICU reasoning could yield clinical reasoning skills that generalize beyond critical care. Here we introduce ICU-REACT, a reasoning dataset developed with 19 clinicians through a clinician-in-the-loop framework to teach LLMs to perform information retrieval and context-aware clinical reasoning in the ICU. Using ICU-REACT, we fine-tuned Clin-REACT models spanning 8B-70B parameters and three model families. Across five clinical reasoning benchmarks, Clin-REACT consistently outperformed its backbone models and open-source general-purpose and medical LLMs. Gains extended to different tasks including script concordance tests, and downstream diagnosis and treatment tasks. These findings suggest that expert reasoning supervision in critical care can improve broader clinical reasoning, although prospective evaluation is needed before real-world clinical use.

cs.CL↗

arXiv2026

Risk Stratification for ICU Delirium using Pervasive Ambient Sensing Information

Delirium is a common and serious complication in the Intensive Care Unit (ICU), associated with increased morbidity, prolonged hospital stays, and higher healthcare costs. Despite its prevalence, early prediction and prevention remain challenging. Environmental factors such as ambient sound and light may influence the onset of delirium, yet they are often overlooked in risk assessments. In this study, we examined whether light intensity and sound pressure levels can independently predict delirium across multiple prediction horizons. We evaluated four efficient sequential neural network models on data collected from 9 ICUs across 309 patients to predict delirium for 10 prediction-window sizes. We reported feature importance and direction of influence using Shapley Additive Explanations analysis. The convolutional model achieved the strongest discrimination, with AUC = 0.80 on sound data and on combined data. Sound features were the dominant predictors overall. Integrating sound with light improved short-term ($<1$ week) prediction, with the combined model assigning the highest risk immediately after the sensing period. These findings suggest that passive ambient sensing, especially sound, can add a clinically meaningful, interpretable signal for delirium risk estimation and offer a practical pathway to enrich multimodal ICU prediction and prevention strategies.

cs.LG↗

arXiv2026

A Knowledge-Injection Framework for Zero-Shot Adaptation of LLMs to Delirium Prediction

Large language models show promise for clinical prediction, but zero-shot performance on specialized tasks is limited by incomplete domain knowledge, especially for smaller locally deployable models. We present a lightweight knowledge-injection framework for zero-shot ICU delirium prediction that augments a deterministic natural-language summary of structured electronic health record data with an external clinical knowledge report at inference time, without fine-tuning or retrieval. We evaluate LLaMA 3.1 8B and LLaMA 3.3 70B on 3,160 ICU admissions from the MIMIC IV dataset. Adding a clinically meaningful external knowledge report improves AUROC by 8.57 percentage points for the 8B model and 1.99 percentage points for the 70B model compared to no external knowledge. Relative to a GPT-5.2 frontier-model reference without external knowledge report (AUROC 68.86%), knowledge injection reduces the performance gap from 15.66 to 7.09 AUROC points for LLaMA 8B and from 5.30 to 3.31 AUROC points for LLaMA 70B. Random control reports do not improve performance and often degrade it, indicating that gains depend on clinically meaningful content rather than added prompt length alone. SHAP-based attribution further confirms that the injected knowledge is actively used during prediction. These findings suggest that inference-time knowledge injection can narrow the gap between locally deployable open-weight models and frontier closed models while preserving a practical, privacy-preserving workflow for resource-constrained clinical settings.

cs.CL↗

arXiv2026

Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring

Multimodal large language models (LLMs) are increasingly explored as automated evaluators in clinical settings, yet their scoring behavior on ordinal clinical scales remains poorly understood. We benchmark three frontier LLM families against supervised deep learning models for scoring Clock Drawing Test (CDT) images on two public datasets using the Shulman rubric. While fully fine-tuned Vision Transformers achieve the best calibration (MAE 0.52, within-1 accuracy 91%), zero-shot LLMs remain competitive on tolerance-based agreement (GPT-5 MAE 0.67, within-1 accuracy 92%) despite higher absolute error. However, per-score analysis reveals that all three LLM families exhibit a pronounced central tendency effect (systematic endpoint compression): predictions are systematically compressed toward the middle of the scale, with over-prediction at the low end (score 0 to 1) and under-prediction at the high end (score 5 to 4). This effect disproportionately affects the clinically critical extremes where accurate scoring most impacts screening decisions for cognitive impairment. Targeted ablations show that neither few-shot exemplars spanning the full score range nor removing clinical terminology from the prompt eliminates the effect. Our findings extend the LLM-as-a-judge bias literature from NLP evaluation to clinical assessment, and highlight the need for calibration-aware evaluation and post-hoc calibration before deploying LLM-based raters in high-stakes screening workflows.

cs.CV↗

arXiv2025

MELON: Multimodal Mixture-of-Experts with Spectral-Temporal Fusion for Long-Term Mobility Estimation in Critical Care

Patient mobility monitoring in intensive care is critical for ensuring timely interventions and improving clinical outcomes. While accelerometry-based sensor data are widely adopted in training artificial intelligence models to estimate patient mobility, existing approaches face two key limitations highlighted in clinical practice: (1) modeling the long-term accelerometer data is challenging due to the high dimensionality, variability, and noise, and (2) the absence of efficient and robust methods for long-term mobility assessment. To overcome these challenges, we introduce MELON, a novel multimodal framework designed to predict 12-hour mobility status in the critical care setting. MELON leverages the power of a dual-branch network architecture, combining the strengths of spectrogram-based visual representations and sequential accelerometer statistical features. MELON effectively captures global and fine-grained mobility patterns by integrating a pre-trained image encoder for rich frequency-domain feature extraction and a Mixture-of-Experts encoder for sequence modeling. We trained and evaluated the MELON model on the multimodal dataset of 126 patients recruited from nine Intensive Care Units at the University of Florida Health Shands Hospital main campus in Gainesville, Florida. Experiments showed that MELON outperforms conventional approaches for 12-hour mobility status estimation with an overall area under the receiver operating characteristic curve (AUROC) of 0.82 (95\%, confidence interval 0.78-0.86). Notably, our experiments also revealed that accelerometer data collected from the wrist provides robust predictive performance compared with data from the ankle, suggesting a single-sensor solution that can reduce patient burden and lower deployment costs...

cs.LG↗

arXiv2025

MANDARIN: Mixture-of-Experts Framework for Dynamic Delirium and Coma Prediction in ICU Patients: Development and Validation of an Acute Brain Dysfunction Prediction Model

Acute brain dysfunction (ABD) is a common, severe ICU complication, presenting as delirium or coma and leading to prolonged stays, increased mortality, and cognitive decline. Traditional screening tools like the Glasgow Coma Scale (GCS), Confusion Assessment Method (CAM), and Richmond Agitation-Sedation Scale (RASS) rely on intermittent assessments, causing delays and inconsistencies. In this study, we propose MANDARIN (Mixture-of-Experts Framework for Dynamic Delirium and Coma Prediction in ICU Patients), a 1.5M-parameter mixture-of-experts neural network to predict ABD in real-time among ICU patients. The model integrates temporal and static data from the ICU to predict the brain status in the next 12 to 72 hours, using a multi-branch approach to account for current brain status. The MANDARIN model was trained on data from 92,734 patients (132,997 ICU admissions) from 2 hospitals between 2008-2019 and validated externally on data from 11,719 patients (14,519 ICU admissions) from 15 hospitals and prospectively on data from 304 patients (503 ICU admissions) from one hospital in 2021-2024. Three datasets were used: the University of Florida Health (UFH) dataset, the electronic ICU Collaborative Research Database (eICU), and the Medical Information Mart for Intensive Care (MIMIC)-IV dataset. MANDARIN significantly outperforms the baseline neurological assessment scores (GCS, CAM, and RASS) for delirium prediction in both external (AUROC 75.5% CI: 74.2%-76.8% vs 68.3% CI: 66.9%-69.5%) and prospective (AUROC 82.0% CI: 74.8%-89.2% vs 72.7% CI: 65.5%-81.0%) cohorts, as well as for coma prediction (external AUROC 87.3% CI: 85.9%-89.0% vs 72.8% CI: 70.6%-74.9%, and prospective AUROC 93.4% CI: 88.5%-97.9% vs 67.7% CI: 57.7%-76.8%) with a 12-hour lead time. This tool has the potential to assist clinicians in decision-making by continuously monitoring the brain status of patients in the ICU.

cs.AI↗

arXiv2024

MANGO: Multimodal Acuity traNsformer for intelliGent ICU Outcomes

Estimation of patient acuity in the Intensive Care Unit (ICU) is vital to ensure timely and appropriate interventions. Advances in artificial intelligence (AI) technologies have significantly improved the accuracy of acuity predictions. However, prior studies using machine learning for acuity prediction have predominantly relied on electronic health records (EHR) data, often overlooking other critical aspects of ICU stay, such as patient mobility, environmental factors, and facial cues indicating pain or agitation. To address this gap, we present MANGO: the Multimodal Acuity traNsformer for intelliGent ICU Outcomes, designed to enhance the prediction of patient acuity states, transitions, and the need for life-sustaining therapy. We collected a multimodal dataset ICU-Multimodal, incorporating four key modalities, EHR data, wearable sensor data, video of patient's facial cues, and ambient sensor data, which we utilized to train MANGO. The MANGO model employs a multimodal feature fusion network powered by Transformer masked self-attention method, enabling it to capture and learn complex interactions across these diverse data modalities even when some modalities are absent. Our results demonstrated that integrating multiple modalities significantly improved the model's ability to predict acuity status, transitions, and the need for life-sustaining therapy. The best-performing models achieved an area under the receiver operating characteristic curve (AUROC) of 0.76 (95% CI: 0.72-0.79) for predicting transitions in acuity status and the need for life-sustaining therapy, while 0.82 (95% CI: 0.69-0.89) for acuity status prediction...

eess.SP↗

arXiv2024

Enhancing EHR Systems with data from wearables: An end-to-end Solution for monitoring post-Surgical Symptoms in older adults

Mobile health (mHealth) apps have gained popularity over the past decade for patient health monitoring, yet their potential for timely intervention is underutilized due to limited integration with electronic health records (EHR) systems. Current EHR systems lack real-time monitoring capabilities for symptoms, medication adherence, physical and social functions, and community integration. Existing systems typically rely on static, in-clinic measures rather than dynamic, real-time patient data. This highlights the need for automated, scalable, and human-centered platforms to integrate patient-generated health data (PGHD) within EHR. Incorporating PGHD in a user-friendly format can enhance patient symptom surveillance, ultimately improving care management and post-surgical outcomes. To address this barrier, we have developed an mHealth platform, ROAMM-EHR, to capture real-time sensor data and Patient Reported Outcomes (PROs) using a smartwatch. The ROAMM-EHR platform can capture data from a consumer smartwatch, send captured data to a secure server, and display information within the Epic EHR system using a user-friendly interface, thus enabling healthcare providers to monitor post-surgical symptoms effectively.

q-bio.QM↗

arXiv2024

DeLLiriuM: A large language model for delirium prediction in the ICU using structured EHR

Delirium is an acute confusional state that has been shown to affect up to 31% of patients in the intensive care unit (ICU). Early detection of this condition could lead to more timely interventions and improved health outcomes. While artificial intelligence (AI) models have shown great potential for ICU delirium prediction using structured electronic health records (EHR), most of them have not explored the use of state-of-the-art AI models, have been limited to single hospitals, or have been developed and validated on small cohorts. The use of large language models (LLM), models with hundreds of millions to billions of parameters, with structured EHR data could potentially lead to improved predictive performance. In this study, we propose DeLLiriuM, a novel LLM-based delirium prediction model using EHR data available in the first 24 hours of ICU admission to predict the probability of a patient developing delirium during the rest of their ICU admission. We develop and validate DeLLiriuM on ICU admissions from 104,303 patients pertaining to 195 hospitals across three large databases: the eICU Collaborative Research Database, the Medical Information Mart for Intensive Care (MIMIC)-IV, and the University of Florida Health's Integrated Data Repository. The performance measured by the area under the receiver operating characteristic curve (AUROC) showed that DeLLiriuM outperformed all baselines in two external validation sets, with 0.77 (95% confidence interval 0.76-0.78) and 0.84 (95% confidence interval 0.83-0.85) across 77,543 patients spanning 194 hospitals. To the best of our knowledge, DeLLiriuM is the first LLM-based delirium prediction tool for the ICU based on structured EHR data, outperforming deep learning baselines which employ structured features and can provide helpful information to clinicians for timely interventions.

cs.AI↗

arXiv2024

Leveraging Computer Vision in the Intensive Care Unit (ICU) for Examining Visitation and Mobility

Despite the importance of closely monitoring patients in the Intensive Care Unit (ICU), many aspects are still assessed in a limited manner due to the time constraints imposed on healthcare providers. For example, although excessive visitations during rest hours can potentially exacerbate the risk of circadian rhythm disruption and delirium, it is not captured in the ICU. Likewise, while mobility can be an important indicator of recovery or deterioration in ICU patients, it is only captured sporadically or not captured at all. In the past few years, the computer vision field has found application in many domains by reducing the human burden. Using computer vision systems in the ICU can also potentially enable non-existing assessments or enhance the frequency and accuracy of existing assessments while reducing the staff workload. In this study, we leverage a state-of-the-art noninvasive computer vision system based on depth imaging to characterize ICU visitations and patients' mobility. We then examine the relationship between visitation and several patient outcomes, such as pain, acuity, and delirium. We found an association between deteriorating patient acuity and the incidence of delirium with increased visitations. In contrast, self-reported pain, reported using the Defense and Veteran Pain Rating Scale (DVPRS), was correlated with decreased visitations. Our findings highlight the feasibility and potential of using noninvasive autonomous systems to monitor ICU patients.

cs.CV↗

arXiv2023

Detecting Visual Cues in the Intensive Care Unit and Association with Patient Clinical Status

Intensive Care Units (ICU) provide close supervision and continuous care to patients with life-threatening conditions. However, continuous patient assessment in the ICU is still limited due to time constraints and the workload on healthcare providers. Existing patient assessments in the ICU such as pain or mobility assessment are mostly sporadic and administered manually, thus introducing the potential for human errors. Developing Artificial intelligence (AI) tools that can augment human assessments in the ICU can be beneficial for providing more objective and granular monitoring capabilities. For example, capturing the variations in a patient's facial cues related to pain or agitation can help in adjusting pain-related medications or detecting agitation-inducing conditions such as delirium. Additionally, subtle changes in visual cues during or prior to adverse clinical events could potentially aid in continuous patient monitoring when combined with high-resolution physiological signals and Electronic Health Record (EHR) data. In this paper, we examined the association between visual cues and patient condition including acuity status, acute brain dysfunction, and pain. We leveraged our AU-ICU dataset with 107,064 frames collected in the ICU annotated with facial action units (AUs) labels by trained annotators. We developed a new "masked loss computation" technique that addresses the data imbalance problem by maximizing data resource utilization. We trained the model using our AU-ICU dataset in conjunction with three external datasets to detect 18 AUs. The SWIN Transformer model achieved 0.57 mean F1-score and 0.89 mean accuracy on the test set. Additionally, we performed AU inference on 634,054 frames to evaluate the association between facial AUs and clinically important patient conditions such as acuity status, acute brain dysfunction, and pain.

cs.CV↗

arXiv2023

Transformers in Healthcare: A Survey

With Artificial Intelligence (AI) increasingly permeating various aspects of society, including healthcare, the adoption of the Transformers neural network architecture is rapidly changing many applications. Transformer is a type of deep learning architecture initially developed to solve general-purpose Natural Language Processing (NLP) tasks and has subsequently been adapted in many fields, including healthcare. In this survey paper, we provide an overview of how this architecture has been adopted to analyze various forms of data, including medical imaging, structured and unstructured Electronic Health Records (EHR), social media, physiological signals, and biomolecular sequences. Those models could help in clinical diagnosis, report generation, data reconstruction, and drug/protein synthesis. We identified relevant studies using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We also discuss the benefits and limitations of using transformers in healthcare and examine issues such as computational cost, model interpretability, fairness, alignment with human values, ethical implications, and environmental impact.

cs.AI↗

arXiv2023

AI-Enhanced Intensive Care Unit: Revolutionizing Patient Care with Pervasive Sensing

The intensive care unit (ICU) is a specialized hospital space where critically ill patients receive intensive care and monitoring. Comprehensive monitoring is imperative in assessing patients conditions, in particular acuity, and ultimately the quality of care. However, the extent of patient monitoring in the ICU is limited due to time constraints and the workload on healthcare providers. Currently, visual assessments for acuity, including fine details such as facial expressions, posture, and mobility, are sporadically captured, or not captured at all. These manual observations are subjective to the individual, prone to documentation errors, and overburden care providers with the additional workload. Artificial Intelligence (AI) enabled systems has the potential to augment the patient visual monitoring and assessment due to their exceptional learning capabilities. Such systems require robust annotated data to train. To this end, we have developed pervasive sensing and data processing system which collects data from multiple modalities depth images, color RGB images, accelerometry, electromyography, sound pressure, and light levels in ICU for developing intelligent monitoring systems for continuous and granular acuity, delirium risk, pain, and mobility assessment. This paper presents the Intelligent Intensive Care Unit (I2CU) system architecture we developed for real-time patient monitoring and visual assessment.

cs.AI↗

arXiv2023

Predicting risk of delirium from ambient noise and light information in the ICU

Existing Intensive Care Unit (ICU) delirium prediction models do not consider environmental factors despite strong evidence of their influence on delirium. This study reports the first deep-learning based delirium prediction model for ICU patients using only ambient noise and light information. Ambient light and noise intensities were measured from ICU rooms of 102 patients from May 2021 to September 2022 using Thunderboard, ActiGraph sensors and an iPod with AudioTools application. These measurements were divided into daytime (0700 to 1859) and nighttime (1900 to 0659). Deep learning models were trained using this data to predict the incidence of delirium during ICU stay or within 4 days of discharge. Finally, outcome scores were analyzed to evaluate the importance and directionality of every feature. Daytime noise levels were significantly higher than nighttime noise levels. When using only noise features or a combination of noise and light features 1-D convolutional neural networks (CNN) achieved the strongest performance: AUC=0.77, 0.74; Sensitivity=0.60, 0.56; Specificity=0.74, 0.74; Precision=0.46, 0.40 respectively. Using only light features, Long Short-Term Memory (LSTM) networks performed best: AUC=0.80, Sensitivity=0.60, Specificity=0.77, Precision=0.37. Maximum nighttime and minimum daytime noise levels were the strongest positive and negative predictors of delirium respectively. Nighttime light level was a stronger predictor of delirium than daytime light level. Total influence of light features outweighed that of noise features on the second and fourth day of ICU stay. This study shows that ambient light and noise intensities are strong predictors of long-term delirium incidence in the ICU. It reveals that daytime and nighttime environmental factors might influence delirium differently and that the importance of light and noise levels vary over the course of an ICU stay.

cs.LG↗

arXiv2022

End-to-End Machine Learning Framework for Facial AU Detection in Intensive Care Units

Pain is a common occurrence among patients admitted to Intensive Care Units. Pain assessment in ICU patients still remains a challenge for clinicians and ICU staff, specifically in cases of non-verbal sedated, mechanically ventilated, and intubated patients. Current manual observation-based pain assessment tools are limited by the frequency of pain observations administered and are subjective to the observer. Facial behavior is a major component in observation-based tools. Furthermore, previous literature shows the feasibility of painful facial expression detection using facial action units (AUs). However, these approaches are limited to controlled or semi-controlled environments and have never been validated in clinical settings. In this study, we present our Pain-ICU dataset, the largest dataset available targeting facial behavior analysis in the dynamic ICU environment. Our dataset comprises 76,388 patient facial image frames annotated with AUs obtained from 49 adult patients admitted to ICUs at the University of Florida Health Shands hospital. In this work, we evaluated two vision transformer models, namely ViT and SWIN, for AU detection on our Pain-ICU dataset and also external datasets. We developed a completely end-to-end AU detection pipeline with the objective of performing real-time AU detection in the ICU. The SWIN transformer Base variant achieved 0.88 F1-score and 0.85 accuracy on the held-out test partition of the Pain-ICU dataset.

cs.CV↗

arXiv2022

Automatic Ultrasound Image Segmentation of Supraclavicular Nerve Using Dilated U-Net Deep Learning Architecture

Automated object recognition in medical images can facilitate medical diagnosis and treatment. In this paper, we automatically segmented supraclavicular nerves in ultrasound images to assist in injecting peripheral nerve blocks. Nerve blocks are generally used for pain treatment after surgery, where ultrasound guidance is used to inject local anesthetics next to target nerves. This treatment blocks the transmission of pain signals to the brain, which can help improve the rate of recovery from surgery and significantly decrease the requirement for postoperative opioids. However, Ultrasound Guided Regional Anesthesia (UGRA) requires anesthesiologists to visually recognize the actual nerve position in the ultrasound images. This is a complex task given the myriad visual presentations of nerves in ultrasound images, and their visual similarity to many neighboring tissues. In this study, we used an automated nerve detection system for the UGRA Nerve Block treatment. The system can recognize the position of the nerve in ultrasound images using Deep Learning techniques. We developed a model to capture features of nerves by training two deep neural networks with skip connections: two extended U-Net architectures with and without dilated convolutions. This solution could potentially lead to an improved blockade of targeted nerves in regional anesthesia.

eess.IV↗

arXiv2020

Facial Action Unit Detection on ICU Data for Pain Assessment

Current day pain assessment methods rely on patient self-report or by an observer like the Intensive Care Unit (ICU) nurses. Patient self-report is subjective to the individual and suffers due to poor recall. Pain assessment by manual observation is limited by the number of administrations per day and staff workload. Previous studies showed the feasibility of automatic pain assessment by detecting Facial Action Units (AUs). Pain is observed to be associated with certain facial action units (AUs). This method of pain assessment can overcome the pitfalls of present-day pain assessment techniques. All the previous studies are limited to controlled environment data. In this study, we evaluated the performance of OpenFace an open-source facial behavior analysis tool and AU R-CNN on the real-world ICU data. Presence of assisted breathing devices, variable lighting of ICUs, patient orientation with respect to camera significantly affected the performance of the models, although these showed the state-of-the-art results in facial behavior analysis tasks. In this study, we show the need for automated pain assessment system which is trained on real-world ICU data for clinically acceptable pain assessment system.

cs.CV↗