Ahmet Kursat Karaman

Objective: To evaluate the diagnostic performance of 2 multimodal large language models (LLMs), ChatGPT-4V and Claude Sonnet 3.7, in preoperative magnetic resonance imaging (MRI)-based diagnosis and histopathological subtyping of WHO Grade I meningiomas and to compare their performance with a radiology resident. Methods: This retrospective study included 113 patients with WHO Grade I meningiomas who underwent preoperative MRI between January 2020 and December 2024. Multisequence MRI (T1WI, T2WI, FLAIR, T1WI+C, DWI, ADC maps) and radiologist-extracted imaging features were provided to both LLMs in a structured "images + human extracted features" format. Each evaluator performed tumor type and subtype prediction using the 2021 WHO CNS classification. Diagnostic accuracies with 95% confidence intervals were calculated, and comparisons between LLMs and the resident were performed using McNemar's test. Results: For type classification, ChatGPT-4V achieved 76.99% accuracy and Claude Sonnet 3.7 achieved 73.45%, while the radiology resident showed significantly higher accuracy at 92.04% (P < .001 for both). Subtype classification was challenging. ChatGPT-4V achieved 20.35% accuracy, Claude Sonnet 3.7 30.09%, and the resident 30.97%, with no statistically significant differences among them. No significant performance differences were observed between the 2 LLMs. Conclusion: Multimodal LLMs demonstrated moderate ability to identify WHO Grade I meningiomas but did not match the diagnostic performance of the radiology resident. Subtype prediction remained limited, reflecting the constraints of conventional MRI for distinguishing subtle histopathological subtypes. Large language models may assist preoperative assessment, but expert interpretation remains essential. Further development using larger datasets and domain-specific model refinement may improve clinical usefulness. Cite this article as: Karaman AK, Demircan NG, Yılmaz B, et al. Preoperative MRI-based diagnosis and subtyping of WHO grade I meningiomas using multimodal large language models: A comparative reader study. Cerrahpaşa Med J. 2026, 50, 0003, doi: 10.5152/cjm.2026.26003.

This study aimed to assess the diagnostic accuracy of three large language models (LLMs)-ChatGPT-4o, Grok-3, and Claude 3.7 Sonnet-in differentiating cerebellopontine angle (CPA) schwannomas and meningiomas using clinical data combined with either raw images or structured imaging features. This retrospective pilot study included 53 patients with pathologically confirmed CPA tumors (28 meningiomas, 25 schwannomas). Each case was submitted to LLMs with clinical data and either raw imaging slices (clinical + images), expert-derived structured imaging features (clinical + expert prompt), or resident-derived structured features (clinical + resident prompt). To establish task difficulty and provide performance context, traditional machine learning algorithms were trained via nested cross-validation. Two board-certified neuroradiologists (each with 10 years of experience) and two third-year radiology residents independently contributed: one pair (B.K. and B.Y.) generated structured imaging features, while the other pair (S.A. and K.H.O.) independently evaluated all cases for performance comparison. Diagnostic performance was compared using McNemar's and Cochran's Q tests with 95% confidence intervals. With expert-derived structured features, LLMs achieved high accuracy (86.8-94.3%), with GPT-4o reaching 94.3% (95%CI 84.6-98.1%), matching the expert neuroradiologist (92.5%, 95%CI 82.1-97.0%, p = 1.000) and significantly outperforming the resident (79.2%, p = 0.021). Performance was moderate with resident-derived structured features (71.7-73.6%), significantly better than raw image input (p < 0.05), but declined markedly with raw images alone (18.9-52.8%), where all LLMs performed significantly worse than radiologists (p < 0.05). Baseline machine learning models confirmed task feasibility with structured features (AUC-ROC 0.968-0.992 for expert features; 0.669-0.875 for resident features). LLMs, particularly GPT-4o, can achieve expert-level diagnostic performance in CPA tumor differentiation when structured textual imaging features are provided. Performance degrades moderately with resident-derived features and dramatically with raw image input, highlighting the critical importance of structured prompting for clinical LLM applications. Current automated visual reasoning remains markedly limited compared to text-based diagnostic inference.

This study aimed to evaluate the impact of post-COVID syndrome across a multitude of health domains among COVID-19 survivors via comparative analysis in previously hospitalized vs. non-hospitalized patients. A total of 158 adult COVID survivors who had symptoms that continued during post-COVID recovery were included in this prospective observational study. Data on handgrip strength, the 1-minute sit-to-stand test (1-MSTST), radiological scoring (CXR), Medical Research Council (MRC) dyspnea, Fatigue Assessment Scale (FAS), International Physical Activity Questionnaire, Post-COVID-19 Functional Status, Hospital Anxiety-Depression Scale, and Short Form Health Survey 36 (SF-36) health-related quality of life (SF-36 HRQoL) were recorded. Post-COVID 1st-month data revealed dyspnea (MRC scale ≥2 in 57.0% of patients) and fatigue (FAS scores ≥22 in 55.7%), a decrease in handgrip strength (60.0%) and physical activity (60.0%), poor HRQoL (SF-36 general health score ≤50 in 61.4%), and radiological disease progression (CXR score ≥3 in 56.8%) in at least half of patients, along with depression (34.8%) or anxiety (43.7%). Post-COVID 6th-month values for fat-free mass index (p=0.008, respectively), 1-MSTST recovery time (p=0.012), and 3rd-month CXR scores (p=0.024) were significantly higher, and 3rd-month SF-36 social functioning domain scores (p=0.026) were significantly lower in previously hospitalized vs. non-admitted patients. In conclusion, our findings indicate a diverse range of impairments in several health domains in COVID-19 survivors, while radiological disease progression and functional limitations were more remarkable in those with previous hospitalization. A comprehensive assessment of health status and persisting rehabilitation needs is necessary in COVID-19 survivors, regardless of disease severity.

Background/aim: To compare the MRI findings and clinical features of multinodular and vacuolating neuronal tumor (MVNT) and dysembryoplastic neuroepithelial tumor (DNET), and reveal the distinguishing features of these tumors from each other. Materials and methods: Patients with a suspected magnetic resonance imaging (MRI)-based diagnosis of MVNT between 2018 and 2022 were collected from the hospital database. In addition, patients diagnosed with DNET on histopathological examination and who had MRIs in the same time period were included in the study. The MRI findings and clinical features were evaluated for each patient. Results: There were 21 patients in the MVNT group and 20 patients in the DNET group. Headache was the most common symptom in patients with MVNTs (61.9%), whereas seizures were more prevalent in those with DNETs (70%). The most frequent locations for the MVNTs were the frontal and parietal lobes (66.6%), while DNETs were most commonly located in the temporal lobe (60%). All the MVNTs were hyperintense in both fluid-attenuated inversion recovery (FLAIR) and T2-weighted imaging (T2WI). All the DNETs were hyperintense on T2WI. However, on FLAIR, seven (35%) of the DNET lesions were hyperintense, while the remaining 13 lesions showed mixed signal intensity forming a bubbly appearance. Moreover, 20 of 21 (95.23%) MVNTs were hyperintense on diffusion-weighted imaging (DWI) (b800), with no apparent diffusion coefficient hypointensity in the lesions. None of the DNETs showed hyperintensity on DWI. Conclusion: MRI findings, particularly those observed on FLAIR and DWI, may be helpful for distinguishing between MVNTs and DNETs, especially in cases where the differential diagnosis is challenging.

Background: Spontaneous third ventriculostomy (STV) is considered rare in obstructive hydrocephalus cases. Purpose: To investigate the efficacy of phase-contrast magnetic resonance imaging (PC-MRI) and three-dimensional constructive interference in steady-state (3D-CISS) for the diagnosis of STV and the potential changes of various radiological parameters measured in hydrocephalus in cases with STV. Material and Methods: Patients with triventricular hydrocephalus who underwent at least one MRI examination, including PC-MRI and 3D-CISS, between 2010 and 2021 were included. The presence of STV was scored separately on PC-MRI and 3D-CISS as follows: 0 = no evidence of STV; 1 = suspicious findings; and 2 = strong evidence of STV. The total STV score was also created by summing the scores determined on both sequences. Evans’ index, third ventricle diameter (TVD), third ventricle height and floor bowing, and height of the interpeduncular cistern were compared between cases with and without STV. Results: STV diagnosis was confirmed in 10/187 (5.3%) cases. The first observer detected suspicious findings in four and two patients without a confirmed diagnosis on PC-MRI and 3D-CISS, respectively, while the second detected suspicious findings in three and two patients, respectively. There was a strong correlation between PC-MRI, 3D-CISS, total scoring results, and consensus results ( P < 0.001). No significant difference was found in measurements except TVD between the two groups. Conclusion: STV may not be a very rare phenomenon and the combined use of PC-MRI and 3D-CISS for diagnosis may aid in overcoming diagnostic challenges. No significant change can be observed in ventriculomegaly-related parameters in patients with STV.

This case report discusses an 11-month-old male with a dermal sinus tract (DST) complicated by spinal meningitis and a long-segment intramedullary abscess. DSTs are rare congenital malformations that may remain asymptomatic until complicated by infections or neurological deficits. The patient was presented with prolonged fever and motor weakness in the extremities, leading to the discovery of a lumbosacral DST and associated abscess via cerebrospinal fluid (CSF) analysis and contrast-enhanced spinal Magnetic Resonance Imaging (MRI). Prompt surgical excision of the DST, drainage of the abscess, and prolonged antibiotic therapy resulted in significant clinical improvement. This case underscores the importance of early recognition, advanced imaging, and aggressive management to mitigate the severe neurological consequences of DST-associated infections.

Background: Flow-related signal void artifacts can be visualized on the T2-weighted (T2W) three-dimensional sampling perfection with application-optimized contrast (3D-SPACE) sequence. Flow void artifacts in the cerebral aqueduct and the fourth ventricle can provide information about cerebrospinal fluid (CSF) flow dynamics. Purpose: In this study, we aimed to test the performance of the T2W 3D-SPACE sequence in assessing the CSF flow in the aqueduct and/or fourth ventricle. Material and Methods: A total of 137 patients (age range = 3–89 years) who underwent CSF flow study were included. The amount of signal loss on T2W 3D-SPACE due to flow in the aqueduct and fourth ventricle was assessed and graded using a 4-point scale of 0 (absence of flow void) to 3 (signal void filling the aqueduct and entire fourth ventricle). A correlation was then sought between the quantitative values obtained by phase-contrast magnetic resonance imaging (PC-MRI) and the amount of signal void in the 3D-SPACE sequence. Results: At the aqueduct level, there was a statistically significant difference in the forward flow velocity and the flow volume among different grades (all P < 0.001). In the grade 3 group, CSF peak systolic flow velocity and mean flow volume were found to be significantly higher than in the other grades ( P < 0.001). The mean aqueduct area in the grade 0 group was found to be significantly different from that in the other classes ( P < 0.001). Conclusion: The amount of signal loss in the fourth ventricle observed on T2W 3D-SPACE is correlated with the peak systolic velocity and flow volume measured quantitatively in PC-MRI.

Objective To evaluate the clinical characteristics and treatment outcomes of patients with chronic pulmonary aspergillosis (CPA) and to determine risk factors for disease recurrence. Methods A total of 43 patients with CPA (mean ± SD age: 61.4 ± 10.5 years, 83.7% were males) were included in this retrospective study. Data on demographic, clinical and disease–related characteristics, galactomannan (GM) test positivity in bronchoalveolar lavage (BAL) samples, histopathological diagnosis, imaging (CT) findings and CPA forms, antifungal therapy, recurrence rate and time to recurrence were recorded. Results Chronic obstructive pulmonary disease (COPD;76.7%) was the leading predisposing factor, and the aspergillus nodule (37.2%) was the most prevalent CPA form.GM test positivity was noted in 89.7% (35/39) of BAL samples. Median duration of voriconazole treatment was 180 days. CPA recurrence was noted in 14.0% of patients, while the comorbid tuberculosis sequela (66.7% vs. 16.2%, p = 0.02) and mild immunosuppressive disorder (100.0% vs. 51.4%, p = 0.032) were significantly more common in patients with recurrence vs. those without recurrence. Recurrence rate was 50.0% (3 of 6 patients) in patients with simple aspergilloma, and ranged from 0.0% to 25.0% in those with other CPA forms. Treatment duration and time to recurrence ranged 70–270 days and 1.1–37 months, respectively in simple aspergilloma, while they were ranged 150–180 days and 30–43.3 months, respectively in other CPA forms. Conclusions Our findings indicate the importance of considering CPA in differential diagnosis in patients with predisposing conditions, and emphasize the tuberculosis sequela, immunosuppressive disorder and the certain CPA forms managed with shorter duration of antifungal therapy (i.e., simple aspergilloma) as the potential risk factors of CPA recurrence.

Background: Differential diagnosis of hypothalamic-optic chiasmatic gliomas (HOCGs) and craniopharyngiomas on magnetic resonance imaging (MRI) can be quite challenging. Purpose: To compare the MRI features of HOCGs and cranipharyngiomas. Material and Methods: Patients diagnosed with HOCG or craniopharyngioma in histopathological evaluation between 2012 and 2022 and who underwent preoperative contrast-enhanced brain MRI were included. Various MRI features were retrospectively evaluated for each lesion: T2-weighted imaging and fluid attenuation inversion recovery hyperintensity, calcification, cystic change, T1-weighted (T1W) imaging hyperintensity of the cystic component, hemorrhage, involvement of sellar, suprasellar or other adjacent structures, lobulated appearance, presence of hydrocephalus, and contrast enhancement pattern. Apparent diffusion coefficient (ADC) values were also evaluated and compared. Results: Among 38 patients included, 13 (34%) had HOCG and 25 (66%) had craniopharyngioma. Craniopharyngiomas had a significantly higher rate of cystic changes, calcification, and T1W imaging hyperintensity of the cystic component than HOCGs ( P −3mm²/s, P <0.05). Conclusion: Although some neuroimaging findings may overlap, features such as presence of cyst and calcification, brain stem and optic pathway involvement, different enhancement patterns, and ADC values may be helpful in the differential diagnosis of HOCGs and craniopharyngiomas.