Skip to content
Log in
Sponsored Have an interesting project or opportunity? → Hire me ↗
Sponsored Artificial Intelligence: A Modern Approach, Global Edition Artificial Intelligence: A Modern Approach, Global Edition

Author

N. Tan

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Open access Aug 2026

Frequency of anti‐neural antibodies and autoimmune epilepsy in focal epilepsy of unknown etiology: An observational study in a Singaporean cohort

Abstract Objective Autoimmune epilepsy (AES) is increasingly recognized as a condition in patients with epilepsy of unknown etiology. Early immunotherapy improves outcomes; however, data on its prevalence and the frequency of anti‐neural/neuronal antibodies in Asian populations remain scarce. Herein, we aim to (1) evaluate the frequency of serum anti‐neuronal/neural antibodies, (2) compare the clinical characteristics between seropositive and seronegative patients, and (3) explore the clinical utility of the ACES (Antibodies Contributing to Focal Epilepsy Signs and Symptoms) score in diagnosing AES in a cohort of Singaporean patients with focal epilepsy of unknown etiology. Methods This was a cross‐sectional, observational, single‐center study performed at the National Neuroscience Institute (NNI), Singapore. Patients with focal epilepsy of unknown etiology were prospectively recruited. Serum samples were tested using both cell‐based (CBA) and immunohistochemistry (IHC)‐based tissue assays. Patients were classified into definite, probable, and possible AES based on antibody positivity. Clinical characteristics, MRI and EEG findings, and ACES scores were analyzed. Results One‐hundred‐and‐twenty patients were included. Nine patients (7.50%) were seropositive on IHC, while 2 patients (1.67%) were positive on CBA (1 with concomitant IHC positivity, 1 without). In total, 10 patients (8.33%) were seropositive on IHC and/or CBA. Seropositivity was associated with female sex, autoimmune and psychiatric comorbidities, behavioral changes, speech problems, abnormal MRI brain, medial temporal T2/FLAIR hyperintensities, and ACES ≥2. AUC of ACES for seropositivity was 0.878; sensitivity and specificity were 60.0% and 96.4% respectively at a cutoff of ≥2. Significance The frequency of specific, well‐characterized anti‐neural/neuronal antibodies in focal epilepsy of unknown etiology is low in Singapore. Distinct clinico‐radiological features, in particular those that comprise the ACES score, were useful in identifying AES. Their presence should prompt antibody testing to confirm the diagnosis of AES for earlier immunotherapy. Plain Language Summary This study performed in Singapore investigated autoimmune epilepsy in people who have focal epilepsy with no clear cause. Researchers tested blood samples for antibodies linked to autoimmune epilepsy and compared clinical features between patients with and without these antibodies. Out of 120 patients, 10 had positive antibody tests—they were more often female, more likely to have autoimmune or psychiatric conditions, behavior or speech problems, and abnormal brain MRI findings. An ACES score of 2 or more was useful for identifying patients with antibodies. Using these clinical features can help clinicians decide who should be tested and treated earlier.

Seong Jin Park, Jeanne May May Tan, Xuejuan Peng et al. · 0 citations
Jun 2026

CLExEval: A Human-in-the-Loop Framework for Qualitative Evaluation of LLM Clinical Reasoning

Large Language Models (LLMs) achieve strong results on many medical benchmarks, but their clinical reasoning remains difficult to evaluate reliably. A central risk is an evaluation illusion: fluent and well-structured explanations can appear clinically convincing even when the final diagnosis is incorrect. We introduce CLExEval, a human-in-the-loop framework for evaluating LLM clinical reasoning under progressive information masking. CLExEval combines 5,600 expert-physician annotations with 200 clinical reasoning traces derived from 40 rare diagnostic cases. Our analysis identifies three recurring failure patterns: (i) verbosity bias, where GPT-4o-mini's diagnostic accuracy drops from 95.0% to 32.5% under information scarcity; (ii) a hidden knowledge paradox, where a specialist model reaches 92.5% maximum diagnostic potential but fails to retrieve that knowledge reliably in verbose contexts; and (iii) a 68.6% reasoning-to-output mismatch, where correct diagnoses appear in reasoning traces but are not reflected in final answers. We further evaluate the LLM-as-a-Judge paradigm on a human-verified failure set (n = 142). GPT-4o-mini approved 47.9% of clinically incorrect outputs, while HuatuoGPT-o1 approved all validly scored failures and showed a positive self-preference bias. These results suggest that standalone automated clinical evaluations can substantially overestimate clinical reliability without expert-grounded validation.

Abin Roy, Afthab Salam Kanniyan, Jawadh Abdul Kabeer et al. · 0 citations