Qi Yong Hemis Ai

Objectives: MRI is used for staging head and neck cancer (HNC), but assigning T- and N-category criteria requires specialised expertise, and so many institutions offer only descriptive MRI reports. This study assessed potentials of commercial large language models (LLMs) to stage oral cavity cancer (OCC) using descriptive MRI reports and compared their accuracy with human experts. Materials and methods: 104 eligible MRI reports were processed by five commercial LLMs (ChatGPT5.4, ChatGPT5.0, ChatGPT4.1, Gemini3.1 and DeepSeekV3.2). T- and N-categories, and overall stage were extracted from outputs of the LLMs. MRI reports were also staged by two multidisciplinary team (MDT) members. Accuracies of the LLMs and MDT members for cancer staging were assessed against the reference staging standard and compared using the McNemar test. Results: The LLMs showed accuracy of 54.8–76.0% for T-categorisation, 63.5–88.5% for N-categorisation, and 53.8–80.8% for overall stage. Compared with DeepSeekV3.2, ChatGPT and Gemini3.1 showed significantly higher accuracy ( p ≤ 0.001), except for Gemini3.1 for T-categorisation ( p = 0.07). No differences in accuracy for staging between ChatGPT versions and between them and Gemini3.1 ( p = 0.08 to > 0.99). MDT members showed accuracy of 74.0–76.9% for T-categorisation, 76.9–77.9% for N-categorisation and 68.3–71.2% for overall stage. The tested LLMs did not consistently outperform MDT members for staging. Conclusion: Variable tested LLM performance in staging OCC from descriptive MRI reports suggested that they are not suitable as standalone staging tools. Additionally, Low MDT staging accuracy highlighted the need for structured reports that include clinical cancer staging to facilitate MDT assessment, thus potentially ensuring optimised disease management. Key Points: Question Can commercial LLMs accurately assign cancer stage based on descriptive MRI reports to overcome the lack of specialised expertise in clinical practice? Findings Tested LLMs demonstrated variable performances, ranging 54.8–76.0% for T-categorisation, 63.5–88.5% for N-categorisation and 53.8–80.8% for overall stage, and failed to consistently outperform MDT members. Critical relevance statement Variable tested LLM performance in staging OCC from descriptive MRI reports suggested that they are not suitable as standalone staging tools. MDT members’ low performances highlight that structured MRI reports with clinical cancer staging are needed to improve multidisciplinary assessment.

Purposes: Habitat imaging analysis has recently been applied to head and neck cancer (HNC) imaging research for various clinical applications. This review systematically evaluated current evidence on its performance in various HNC applications, highlighting its advantages and limitations. Methods: A systematic search of PubMed, Web of Science, and Scopus was conducted for articles from database inception to 30 January 2026, using terms related to habitat analysis, HNC, and medical imaging. Patient numbers, tumour origin, imaging modality, clustering features, habitat methods, subregion numbers, clinical applications, and performance of the models were extracted from the included articles. Results: Twenty-five original articles published between 2019 and 2026 were eligible for analysis, of which 10 had external validation datasets. Nasopharyngeal carcinoma was the most studied HNC ( n = 9); CT was the most used modality ( n = 9); and K-means clustering was the predominant habitat method ( n = 20). Habitat imaging analysis performed consistently well and outperformed conventional radiomics or clinical models in predicting pathological/immunohistochemical (IHC) biomarkers, cancer characterisation and nodal status, with areas under the curve (AUCs) of > 0.8, but showed variable results for treatment response prediction. Combining the multiple models slightly improved performances in most studies. Conclusion: Habitat imaging analysis showed potentials in various applications in HNC, performing consistently well in predicting pathological/IHC biomarker, tumour characterisation and nodal status, but not in treatment response prediction. Combining multiple models slightly improved performance, raising questions about whether such complex integrations offer clinical value. Current evidence remains heterogeneous, and so prospectively conducted and externally validated studies are required to clarify their practical utility.

Objectives: To compare the visibility of the mandibular canal (MC) at three mesial-distal levels adjacent to the mandibular third molar (M3M) using standard-dose and four low-dose cone-beam computed tomography (CBCT) protocols, and evaluate their impact on assessing the spatial relationship between the MC and M3M. Methods: A total of 100 CBCT scans were obtained from 20 dry mandibles with M3Ms, using one standard-dose, normal-resolution protocol (0.2 mm voxel; DAP: 322 mGy×cm ² ) and four low-dose protocols with varying resolutions (0.15-0.4 mm voxel; DAP: 23-131 mGy×cm ² ). For each scan, MC visibility at the distal, middle, and mesial thirds of the M3M was rated on a 4-point scale. The spatial relationship between the MC and M3M apex was evaluated both qualitatively (non-contact, contact, or penetration) and quantitatively (the MC-M3M distance). Pairwise comparisons between the standard-dose and each low-dose protocol were conducted using the Wilcoxon signed-rank test. Results: The MC visibility at all levels was high (≥3.4/4) for all protocols except the low-dose, low-resolution protocol (voxel size: 0.4 mm; DAP: 23 mGy×cm ² ). Significantly lower MC visibility was observed only with the low-dose, low-resolution protocol compared to the standard-dose protocol ( p = 0.006-0.026). No significant differences were found in the qualitative and quantitative assessments of the spatial relationship between the MC and M3M apex among any of the protocol pairs. Conclusions: Although the lowest-dose, lowest-resolution CBCT reduced MC visibility, no significant differences were observed between standard-dose and low-dose protocols in assessing the MC-M3M apex relationship. Notably, the lack of significant difference should not be interpreted as evidence of equivalence, as the study used subjective image assessments without an independent reference standard. Low-dose CBCT may be considered in selected cases, but protocol selection should remain indication-oriented, patient-specific, and device-specific.

Background/Objectives: Corpus callosum atrophy is a well recognized feature of multiple sclerosis (MS), which has been associated with disease duration and severity. Quantitative T1rho imaging is an MRI technique sensitive to microstructural tissue alterations, but the T1rho characteristics of the corpus callosum in MS remain largely unexplored. This study aimed to characterize T1rho in the corpus callosum in MS patients, and examine relationships between T1rho values, callosal atrophy, and clinical and imaging markers of disease burden. Method: Thirty patients with relapsing-remitting multiple sclerosis (RRMS) and 30 healthy controls underwent T1rho imaging. T1rho values were measured in the genu, body, splenium, and whole corpus callosum. Comparisons between patient and control groups were performed using the Mann–Whitney U test and the independent samples t-test, and Pearson’s and Spearman’s correlation coefficients were used to evaluate the correlation between imaging and clinical measures. Results: T1rho values were significantly increased across all corpus callosum subregions in patients compared to controls. Corpus callosum T1rho values showed correlations with disease duration (r = 0.375–0.408, p < 0.041), lesion load (r = 0.726–0.810, p < 0.001), and the corpus callosum index (CCI) (r = −0.731–−0.640, p < 0.001). Conclusions: T1rho values are elevated in the corpus callosum of MS patients and associated with disease duration, lesion load, and callosal atrophy.

Background: Nasopharyngeal carcinoma (NPC) can be detected early on MRI, but adoption for screening is limited by a shortage of experienced specialists. MRI artificial intelligence (AI) algorithms for diagnosing NPC in the literature may address this, but most studies focus on routine clinical care, and their applicability in screening requires close examination. Purpose: To (i) evaluate the diagnostic performance of AI for NPC detection, (ii) analyze the impact of study protocols, and (iii) assess the applicability of existing studies to NPC screening. Study Type: Systematic review. Population: Thirty‐eight studies were included, including a total of 23,398 patients (20,693 NPC and 2705 non‐NPC). Fieldstrength: 1.5 T to 3.0 T. Assessment: Following PRISMA guidelines, PubMed, Scopus, and Embase records from January 1, 2009 to August 16, 2025 were screened by two independent reviewers for studies on NPC detection, localization, and/or diagnosis using MRI. Included studies were reviewed for applicability and risk of bias. Lesion‐localization performance (Dice similarity score [DSC]) and NPC vs. non‐NPC discrimination (sensitivity/specificity) were extracted and summarized. Subgroup analyses assessed the impact of intravenous contrast on AI performance. Statistical Tests: Meta‐analysis used standard random‐effects univariate restricted maximum likelihood model and hierarchical summary receiver operator characteristics for performance pooling and subgroup analysis, taking p value < 0.05 as significance. Results: Pooled performance from 30 localization and 6 discrimination studies was DSC = 0.80 (CI: 0.78–0.82) and sensitivity/specificity = 97.1% (CI: 88.0%–99.3%)/87.8% (CI: 78.9%–93.2%), respectively. Intravenous contrast did not significantly affect AI performance for either task ( p > 0.05). Only seven studies were fully applicable to screening. Data Conclusion: AI‐driven NPC screening using MRI is feasible at high sensitivity. Using non‐contrast MRI did not significantly impact AIs' performance. However, applicability and the suboptimal ratio of NPC: non‐NPC patients remain the weaknesses of existing studies, warranting further investigation with cohorts that better reflect the screening scenario. Systematic Review Registration: The systematic review protocol of this system is registered with the Inplasy registry (Ref. INPLASY202570076).

Objectives: This study aimed to identify and synthesize imaging findings in the temporomandibular joint (TMJ) complex following radiation therapy (RT) for head and neck cancer (HNC). Materials and methods: A systematic search of five databases retrieved studies reporting post-RT imaging of TMJ structures. Extracted data included patient demographics, RT dose and technique, imaging modality, imaging changes and clinical findings. Risk of bias was assessed using ROBINS-I V2 (Risk of bias in non-randomized studies—of interventions, version 2). Results: Eleven retrospective studies (282 patients; mean age 52.7 years; RT dose 30–110 Gy) were included. Post-RT analysis was evaluated by imaging modalities of MRI (7 studies), CT (4), US (1), and panoramic images (1). The most commonly described structural changes in the muscles of mastication included MRI signal abnormalities (43 patients), atrophy (20 patients) and hypertrophy (20 patients), while condylar changes included erosion (11 patients), surface irregularity (12 patients) or sclerosis (6 patients). Trismus was identified in 126 patients (44.7%). No study described changes to the glenoid fossa or articular eminence or direct osteoradionecrosis (ORN) of TMJ structures. All studies exhibited moderate to critical risk of bias, limiting the certainty of associations. Conclusions: RT can induce structural changes across TMJ components detectable on imaging, which may contribute to trismus and functional impairment. Current evidence is limited, heterogeneous and retrospective, underscoring the need for prospective studies with standardized imaging and clinical correlation. Key Points: Radiation therapy for head and neck cancer can affect jaw structures. Imaging shows changes in muscles and joints that are involved with chewing and mouth opening. These changes may lead to trismus (difficulty opening the mouth). Nearly half of patients in reviewed studies experienced this condition, impacting daily life. Current evidence is limited and inconsistent. Prospective studies with standardized imaging approaches are needed to further clarify radiation-related TMJ changes.

Objectives: This systematic review aimed to investigate the current development and application landscape of generative artificial intelligence (Gen-AI) networks specifically designed for dental and maxillofacial radiology (DMFR) and summarize their potential applications for clinical practice, education, and research. Methods: Five electronic databases were searched to identify studies that developed and validated DMFR-specific Gen-AI networks. Data regarding the purpose and type of the Gen-AI model, dataset details, quantitative evaluation metrics, methods of subjective assessment, and key findings were extracted. Customized assessment criteria adapted from the Checklist for Artificial Intelligence in Medical Imaging (CLAIM) checklist were used to evaluate the risk-of-bias for the included studies in 4 domains (dataset details, reporting of Gen-AI model architecture and training strategies, reliability of performance evaluation methods, and accessibility of code and developed models). Results: Forty-three studies were included from the initially identified 2060 records. Of these, 24 studies (55.8%) focused on image quality, addressing improvements in spatial resolution, artifact and noise reduction, image geometry and projection, quantitative accuracy of voxel values, 14 (32.6%) on image simulation (including the generation of bitewing, panoramic, and cephalometric radiographs, image-to-image translation, and post-treatment prediction simulation), 3 (7%) on 3D reconstruction from 2D images, and 2 (4.6%) on automated interpretation and reporting. Nearly all included studies (95.3%) reported objective evaluation metrics while about half (58.1%) incorporated subjective assessments using scoring systems or visual grading. Risk-of-bias was moderate for dataset details in 7 studies (16.3%) and performance evaluation in 20 (46.5%), and high for code and model accessibility in 35 studies (81.4%). Conclusions: While DMFR-specific Gen-AI models show promising potential for clinical practice, education, and research, their applicability requires overcoming challenges related to data quality, validation, integration, and ethical and legal considerations. Further clinical validation, increased transparency and accessibility, and thorough evaluation of cost-effectiveness in diagnostic workflows are essential to ensure their safe and effective use.

Review Article Radiomic analysis of medical imaging for classification in oncology and recommendations for clinically initiated research: a literature review Kaijing Mao, Lun Matthew Wong, Kuo Feng Hung, Rongli Zhang, Walter Y. H. Lam, Zhiyi Shan, Tiffany Y. So, Qi Yong Hemis Ai, Kyongtae Ty Bae

Purpose: T1rho imaging showed potential applications in cancer imaging but little research explored the underlying biological processes that contribute to the T1rho values in cancer. This study aimed to investigate the potential associations between quantitative imaging biomarkers from T1rho imaging and the well-established diffusion weighted imaging (DWI), with tumour-stromal, immunohistochemical (IHC), and tumour-infiltration-lymphocytes (TIL) biomarkers in nasopharyngeal carcinoma (NPC). Methods: Pre-treatment T1rho and DWI imaging of primary NPCs were performed in 50 prospectively recruited patients. The mean T1rho and apparent diffusion coefficient (ADC) of NPC were obtained and correlated with tumour-stromal, IHC, TIL biomarkers using the Pearson Correlation test and the coefficients (R) were calculated. Results: The mean T1rho values negatively correlated with collagenous stroma-lymphoid stroma (R=-0.314, p = 0.03) and positively correlated with percentage of tumour cells positive for Ki-67 (R = 0.402, p < 0.01), but there were no associations between T1rho values and the other tumour-stromal, IHC or TIL biomarkers (p = 0.16–0.98) or between ADC values and any of these biomarkers (p = 0.07–0.82). Conclusion: Our results showed the possible underlying biological mechanisms of T1rho imaging in head and neck cancer. T1rho imaging negatively correlated with the ratio of collagenous to lymphoid stroma, and positively correlated with tumour cell proliferation, which are both known to be predictors of outcome, suggesting that T1rho imaging may have a valuable role in head and neck cancer imaging. As this is a preliminary study with small sample size, further studies are encouraged to validate our findings.