Panoramic radiography is the primary modality for dental screening, where accurate diagnosis hinges on image quality. However, current deep learning methods often isolate these tasks, limiting the unified reasoning essential for clinical trust. To address this, we present Dental-QAD, a unified multimodal large language model framework designed for joint quality assessment and disease diagnosis. We bridge the scarcity of reasoning data by constructing a large-scale dataset with expert annotations and employing a strategy to synthesize diagnostic reasoning. Furthermore, to strictly align model outputs with clinical standards, we introduce a Multi-Granularity Reinforcement Learning mechanism via Group Relative Policy Optimization that optimizes both continuous quality scores and discrete diagnoses. Our experiments confirm that Dental-QAD surpasses state-of-the-art models in quality correlation and accuracy while offering the interpretable reasoning required for integration into clinical workflows.
This post hoc analysis aimed to explore the effect of perfluorohexyloctane on dry eye disease (DED) associated with meibomian gland dysfunction (MGD) by baseline disease severity. Data from a randomized phase 3 trial (NCT05515471) were divided based on the baseline severity of DED signs (total corneal fluorescein staining [tCFS] score: < 6 or ≥ 6; tear film breakup time: < 3 or ≥ 3; MGD score: < 7 or ≥ 7) or symptoms (eye dryness scores [EDS]: < 70 or ≥ 70; ocular surface disease index score: < 60 or ≥ 60). The main outcomes of interest included response rates at day 57 of tCFS score, EDS, and tCFS and EDS, safety, and tolerability. Participants treated with perfluorohexyloctane had greater odds of achieving EDS (odds ratio [OR]: 2.25; 95
A man in his 80s had been taking atorvastatin for 2 months for hyperlipidaemia. Due to the recurrence of nephrotic syndrome, methylprednisolone, torasemide, benidipine and atorvastatin were prescribed after admission to hospital. While his dyspnoea and oedema improved with diuretic therapy, the patient reported significant muscle soreness 8 days after starting the combined atorvastatin and methylprednisolone. Simultaneously, creatine kinase and myoglobin levels increased significantly. The temporal correlation between the drug combination and symptom onset, coupled with rapid clinical and biochemical improvement after atorvastatin discontinuation, supported a diagnosis of rhabdomyolysis due to a potential drug-drug interaction between atorvastatin and methylprednisolone, likely exacerbated by hypoproteinaemia. This case highlights the need for clinicians and pharmacists to be vigilant for such drug-drug interactions, particularly in patients with predisposing conditions like hypoproteinaemia.
Orthokeratology (Ortho-K) is an important optical intervention for myopia control in adolescents. Its proposed mechanism has evolved from the traditional "peripheral defocus" hypothesis to a complex network of multi-pathway collaborative regulation. By analyzing the latest literature from 2023 to 2025, this review systematically outlines the core mechanisms of Ortho-K in slowing myopia, and innovatively proposes a "Initial driver - Intermediate signal integration - Terminal effect" three-tier regulatory network model based on evidence weight. This model covers: 1) the initial driver tier (the highest evidence weight, including direct lens stimulation on the cornea and peripheral retinal myopic defocus, the core trigger of the regulatory cascade); 2) the intermediate signal integration tier (corneal nerve reflex, retinal-choroidal structural and hemodynamic changes, responsible for signal conversion and amplification); 3) the terminal effect tier (scleral remodeling, the final common pathway of myopia inhibition). The review emphasizes the interaction and prioritization between optical defocus, corneal biomechanics, neurovascular regulation, and molecular signaling pathways, while also assessing the evidence levels of each pathway. Additionally, it addresses key controversies in the current field (such as individual response differences) and explores future research directions in combination with cutting-edge technologies like artificial intelligence (AI) personalized design and single-cell sequencing. Finally, based on mechanistic studies, it provides feasible translational recommendations for optimizing clinical practice and personalized treatment strategies.
This study aimed to elucidate the contemporary clinical characteristics and evolving trends of laryngeal tuberculosis(LTB), thereby providing evidence for optimizing clinical diagnostic and therapeutic strategies. We conducted a retrospective analysis of the clinical data from patients diagnosed with LTB at a single-center medical institution in northern China between July 2014 and June 2025. A total of 49 patients with LTB were included in this study, predominantly males aged 55–64 years. The most common presenting symptom was hoarseness (67.34
