Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Balu Bhasuran has a total of 145 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Probabilistic Coherence of Stated Diagnostic Beliefs in Large Language Models: Anchored Tests on a Synthetic Clinical Corpus

    by Govind Arun, Ajo Babu George, Balu Bhasuran

    Clinical evidence arrives in sequence, yet a coherent diagnostician’s belief should depend on what was found, not on the order of finding it. Bayesian belief revision is order-invariant as a theorem, which yields an exact, annotation-free test, with target zero, of the distribution a clinical language model reports. Three further axioms cover update fidelity, redundancy insensitivity, and positional anchoring. Eleven open-weight configurations of 4B to 32B parameters are evaluated on 1.12 million responses over 1956 patients and 49 pathologies of the synthetic DDXPlus corpus. Every divergence is anchored to the model’s own test–retest floor and between-patient ceiling, and to a corpus replicate floor of 0.049. On the Qwen3 ladder, raw permutation divergence rises from 0.128 to 0.291 with scale, while the order effect as a fraction of each model’s floor-to-ceiling range falls from 0.215 to 0.127. Belief updates do not track the corpus evidence: direction agreement is 0.454 to 0.549, even on findings the corpus moves by more than two log odds. Every model names the correct diagnosis 1.8 to 5.6 times more often than its own stated posterior ranks it first, locating the deficit in the probabilistic interface, not in clinical knowledge.

  • Tutorial: guidance on the use of large language models for medical research

    by Qiao Jin, Nicholas Wan, Robert Leaman, Shubo Tian, Zhizheng Wang, Yifan Yang, Zifeng Wang, Guangzhi Xiong, Po-Ting Lai, Qingqing Zhu, et al.


  • Open Access

    Generating Question Prompt Lists From Electronic Health Record Data Using Large Language Models: Iterative Evaluation Study

    by Zhe He, Balu Bhasuran, Mia Liza A Lustria, Karim Hanna, Michael Killian, Cindy Shavor, Mandy Dailey, Sai Sidharth Manikandan, Xiao Luo

    Background: Patients frequently access laboratory results through patient portals, but many struggle to interpret these values and formulate relevant questions for their clinicians. Question prompt lists (QPLs) can enhance communication but are rarely tailored to individual clinical contexts. Objective: This study evaluated the feasibility of using large language models (LLMs) to generate patient-friendly, clinically relevant questions grounded in electronic health record (EHR) laboratory data. Methods: We extracted deidentified clinical profiles, including laboratory results, diagnoses, and medications, from patients with chronic conditions (eg, diabetes and chronic kidney disease). Using 9 deidentified clinical profiles from the OneFlorida Data Trust, we generated 486 questions across all rounds: 126 with GPT-4o in round 1, 120 with GPT-4o in round 2, 180 with GPT-4o in round 3, and 60 with LLaMA 3.2 in round 3. Prompt refinements were informed by clinician ratings consisting of 2 binary questions (ie, clear phrasing and clinical validity) and 3 Likert-scale questions (ie, significance for the patient’s health; clinical appropriateness, that is, the likelihood of being asked in primary care setting; and willingness to answer). Refinements were incorporated after each round. Patient participants then evaluated selected questions for understandability, perceived usefulness, and intention to use. Readability was assessed with standard indices. Results: Iterative clinician feedback improved question clarity and reduced clinically irrelevant suggestions. Across rounds, GPT-4o consistently produced more coherent and patient-friendly questions, while LLaMA 3.2 demonstrated competitive performance on Likert-scale metrics. It exhibited greater variability in clinical appropriateness as noted by clinicians. In round 3, the binary metric “clear phrasing” reached a ceiling effect for both models, while clinical validity ratings showed greater variability, particularly from one clinician. Likert-scale evaluations tended to favor LLaMA 3.2 across all 3 clinicians for clinical appropriateness (3.37‐4.82 vs 3.02‐4.67), significance for the patient’s health (3.38‐4.28 vs 2.97‐3.83), and willingness to answer (3.17‐4.70 vs 2.82‐4.47), with multiple comparisons reaching statistical significance after Bonferroni correction. Patient evaluation (N=134) of GPT-4o–generated questions showed that 25 of 30 questions had moderate to high understandability (average Likert score ≥3.5), and 19 of 30 questions had moderate to high usefulness (average Likert score ≥3.5). Conclusions: This study supports the feasibility of using LLMs with structured EHR-derived laboratory data to generate contextualized QPLs, but model outputs varied in clinical appropriateness and readability. Clinician-in-the-loop review remains necessary before patient-facing use.


  • Open Access

    A Curated Dataset for Question Answering on Laboratory Test Reference Ranges and Interpretation

    by Balu Bhasuran, Qiao Jin, Angelique Deville, Yonghui Wu, Karim Hanna, Zhiyong Lu, Zhe He

    Laboratory tests are crucial for diagnosing and managing health conditions, offering essential reference ranges for result interpretation. The diversity of lab tests, influenced by variables such as the specimen type (e.g., blood, urine), gender, age, and other influencing factors like pregnancy, makes automated interpretation challenging. Automated clinical decision support systems attempting to interpret these values must account for such nuances to avoid misdiagnoses or incorrect clinical decisions. In this regard, we present LabQAR (Laboratory Question Answering with Reference Ranges), a manually curated dataset comprising 550 lab test reference ranges derived from authoritative medical sources, encompassing 363 unique lab tests and including multiple-choice questions with annotations on reference ranges, specimen types, and other factors impacting interpretation. We also assess the performance of several large language models (LLMs), including LLaMA 3.1, GatorTronGPT, GPT-3.5 turbo, GPT-4 turbo, and GPT-4o, in predicting reference ranges and classifying results as normal, low, or high. The findings indicate that GPT-4o outperforms other models, showcasing the potential of LLMs in clinical decision support.


  • Open Access

    A Unified Multimodal Framework for Dataset Construction and Model-Based Diagnosis of Ameloblastoma

    by Ajo Babu George, Anna Mariam John, Athul Anoop, Balu Bhasuran

    Artificial intelligence (AI)-enabled diagnostics in maxillofacial pathology require structured, high-quality multimodal datasets. However, existing resources provide limited ameloblastoma coverage and lack the format consistency needed for direct model training. We present a newly curated multimodal dataset specifically focused on ameloblastoma, integrating annotated radiological, histopathological, and intraoral clinical images with structured data derived from case reports. Natural language processing techniques were employed to extract clinically relevant features from textual reports, while image data underwent domain specific preprocessing and augmentation. Using this dataset, a multimodal deep learning model was developed to classify ameloblastoma variants, assess behavioral patterns such as recurrence risk, and support surgical planning. The model is designed to accept clinical inputs such as presenting complaint, age, and gender during deployment to enhance personalized inference. Quantitative evaluation demonstrated substantial improvements; variant classification accuracy increased from 46.2 percent to 65.9 percent, and abnormal tissue detection F1-score improved from 43.0 percent to 90.3 percent. Benchmarked against resources like MultiCaRe, this work advances patient-specific decision support by providing both a robust dataset and an adaptable multimodal AI framework.

  • AI for Clinical Reasoning and Diagnosis (Free Seminar)

    by Balu Bhasuran

    This seminar explores methods for integrating laboratory test results, EHR data, causal inference, and large language models to improve diagnostic reasoning and patient-facing interpretation. The content covers representation strategies, evaluation frameworks, prompt and generation constraints, and research practices that strengthen reproducibility and publication potential.

  • Open Access

    AI-Driven Drug Repurposing through miRNA-mRNA Relation

    by Sharanya Manoharan, Balu Bhasuran, Oviya Ramalakshmi Iyyappan, Mohamed Saleem Abdul Shukkoor, Malathi Sellapan, Kalpana Raja

    miRNA mRNA relations are closely linked to several biological processes and disease mechanisms In a recent study we tested the performance of large language models LLMs on extracting miRNA mRNA relations from PubMed PubMedBERT achieved the best performance of 0.783 F1 score for miRNA mRNA Interaction Corpus MMIC Here we first applied the finetuned PubMedBERT model to extract miRNA mRNA relations from PubMed for chronic obstructive pulmonary disease COPD Alzheimers disease AD stroke type 2 diabetes mellitus T2DM chronic liver disease and cancer Next we retrieved miRNA drug relations using KinderMiner a literature mining tool for relation extraction Then we constructed three interaction networks 1 disease centric network 2 drug centric network and 3 miRNA centric network comprising 3497 nodes and 16417 edges organized as a directed graph to capture complex biological relationships Finally we validated the drugs using MIMIC IV Our integrative approach revealed both established and novel candidate drugs for diseases under study through 595 miRNA drug relations extracted from PubMed To the best of our knowledge this is the first study to systematically extract and visualize relationships among four distinct biomedical entities miRNA mRNA drug and disease

  • GRU-CNN Hybrid Model Approach to Classify Human Brain Emotion using EEG Dataset

    by I R Oviya, Balu Bhasuran, Deepak Soni, Rudraksh Mohanty, J C Harshini, Kashish Gurnani, S M Varrshini

    This research investigates the use of deep learning algorithms to classify human emotions from EEG brainwave signals based on two publicly accessible Kaggle datasets. The first dataset involves EEG recordings of one male and one female subject, recorded with a Muse EEG headband on electrode locations TP9, AF7, AF8, and TP10. These signals were preprocessed with statistical feature extraction techniques. The second dataset has self-rated emotional response ratings by 14-16 participants watching 120 one-minute music video clips, rated on arousal, valence, and dominance. It further includes EEG and physiological measures from 32 participants who watched 40 individual videos and rated them. All EEG signals were processed as time-series data and marked as positive, neutral, or negative. To effectively capture both spatial and temporal dynamics of brain activity, a hybrid deep learning model combining Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) was employed. The proposed model achieved an $\mathbf{F 1}$-score of $\mathbf{0. 9 8}$ and an accuracy of $\mathbf{9 8. 8 3} \%$, demonstrating strong emotion recognition performance.


  • Open Access

    Literature-Based Discovery (LBD): Towards Hypothesis Generation and Knowledge Discovery in Biomedical Text Mining

    by Balu Bhasuran, Gurusamy Murugesan, Jeyakumar Natarajan

    Biomedical knowledge is growing at an astounding pace with a majority of this knowledge represented as scientific publications. Text mining tools and methods represent automatic approaches for extracting hidden patterns and trends from this semi-structured and unstructured data. In biomedical text mining, literature-based discovery (LBD) is the process of automatically discovering novel associations between medical terms otherwise mentioned in disjoint literature sets. LBD approaches have proven to successfully reduce the discovery time of potential associations that are hidden in the vast amount of scientific literature. The process focuses on creating concept profiles for medical terms such as a disease or symptom and connecting them with a drug and treatment based on the statistical significance of the shared profiles. This knowledge discovery approach introduced in 1989 remains a core task in text mining. Currently, the ABC principle-based two approaches namely open discovery and closed discovery are mostly explored in the LBD process. This review starts with a general introduction about text mining followed by biomedical text mining followed by a brief introduction of the core ABC principle and its associated two approaches open discovery and closed discovery in the LBD process. This review discusses the deep learning applications in LBD by reviewing the role of transformer models and neural networks-based LBD models and their future aspects. Additionally, the potential of Large Language Models in enriching the LBD process is discussed with challenges and solutions and finally reviews the key biomedical discoveries generated through LBD approaches in biomedicine and concludes with the current limitations and future directions of LBD. Received: 1 February 2025 | Revised: 12 May 2025 | Accepted: 30 May 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study. Author Contribution Statement Balu Bhasuran: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Gurusamy Murugesan: Methodology, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Jeyakumar Natarajan: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration.


  • Open Access

    LabQAR: A Manually Curated Dataset for Question Answering on Laboratory Test Reference Ranges and Interpretation

    by Balu Bhasuran, Qiao Jin, Angelique Deville, Yonghui Wu, Karim Hanna, Zhiyong Lu, Zhe He

    Laboratory tests are crucial for diagnosing and managing health conditions, providing essential reference ranges for result interpretation. The diversity of lab tests, influenced by variables like the specimen type (e.g., blood, urine), gender, age-specific, and other influencing factors such as pregnancy, makes automated interpretation challenging. Automated clinical decision support systems attempting to interpret these values must account for such nuances to avoid misdiagnoses or incorrect clinical decisions. In this regard, we present LabQAR ( Lab oratory Q uestion A nswering with R eference Ranges), a manually curated dataset comprising 550 lab test reference ranges derived from authoritative medical sources, encompassing 363 unique lab tests and including multiple-choice questions with annotations on reference ranges, specimen types, and other factors impacting interpretation. We also assess the performance of several large language models (LLMs), including LLaMA 3.1, GatorTronGPT, GPT-3.5, GPT-4, and GPT-4o, in predicting reference ranges and classifying results as normal, low, or high. The findings indicate that GPT-4o outperforms other models, showcasing the potential of LLMs in clinical decision support.