Japanese Journal of Radiology最新文献

Recent advances in diagnostic imaging have led to an increase in the detection of early-stage breast cancer, prompting growing interest in minimally invasive breast surgery. These approaches include the omission of axillary lymph node dissection, use of endoscopic or robot-assisted surgery, and application of non-surgical ablation techniques. Non-surgical ablation therapy aims to eradicate tumor tissue in situ by delivering localized thermal or cryogenic energy under imaging guidance without surgical resection. Among these, radiofrequency ablation (RFA) and cryoablation have been the most extensively studied, showing promising results for small early-stage tumors. Clinical studies have reported high rates of complete tumor ablation, favorable safety profiles, and excellent cosmetic outcomes. Moreover, single-arm prospective trials have provided 5-year follow-up data, offering valuable insights into long-term efficacy. Despite these encouraging results, non-surgical ablation has not been internationally adopted as a standard treatment because of concerns regarding oncological surveillance and outcome validation. However, in Japan, RFA for early-stage breast cancer was approved for insurance coverage in December 2023. This review summarizes the current trends and clinical evidence for RFA and cryoablation, and discusses the regulatory pathways required for RFA insurance approval in Japan.

Objectives: To evaluate the distribution patterns of sarcoidosis involvement on FDG-PET/CT in patients with known or suspected cardiac sarcoidosis (CS), with a particular focus on upper abdominal lymph nodes (LN) (periportal LN [PLN], anterior pancreaticoduodenal LN [APDLN], and posterior pancreaticoduodenal LN [PPDLN]) and the association of them with other lesions and myocardium.

Methods: We identified 861 FDG-PET/CT scans performed between July 2016 and August 2024 in patients with known or suspected CS, and included 178 cases for analysis of FDG uptake patterns suggestive of sarcoid involvement. FDG-positive LNs or regions were classified as sarcoidosis-related based on treatment response, characteristic uptake patterns, or histological confirmation. The occurrence ratio of FDG-positive lymph nodes or regions was also assessed in relation to myocardial FDG uptake patterns.

Results: FDG uptake was observed most frequently in hilar and mediastinal LNs (79% and 76%, respectively). Upper abdominal LN uptake was observed in 49.4% of patients, most commonly in the PLN (31.5%), APDLN (38.2%), and PPDLN (37.1%). Heatmap analyses revealed strong co-occurrence between thoracic and upper abdominal LNs, suggesting a lymphatic dissemination pattern. Peripheral LNs such as axillary, subclavian, and inguinal/pelvic stations demonstrated low uptake and minimal co-occurrence.

Conclusions: FDG-PET/CT provides valuable insight into the structured lymphatic dissemination of sarcoidosis. Frequent FDG uptake in upper abdominal lymph nodes, particularly when accompanied by thoracic involvement, represents a characteristic finding in sarcoidosis. Recognition of this pattern can improve diagnostic accuracy and help differentiate sarcoidosis from other systemic diseases. This study assessed lymph node involvement on FDG-PET/CT in patients with suspected or known cardiac sarcoidosis, revealing distinct dissemination patterns into the upper abdomen. These findings enhance understanding of disease pathophysiology and may improve diagnostic evaluation.

Purpose: This study aimed to develop and validate machine learning models based on quantitative radiomics parameters extracted from T1-weighted MRI to differentiate enchondromas from atypical cartilaginous tumours (ACTs).

Methods: A retrospective cohort comprising 66 patients (35 with histopathologically confirmed enchondroma and 31 with ACT) was included in the study. T1-weighted MRI images were used for 2D segmentation, performed independently by two experienced observers on all visible slices of each lesion. A comprehensive set of 107 radiomics features was extracted from these segmented regions of interest. LASSO regression was applied for dimensionality reduction. Four distinct machine learning algorithms-Support Vector Machine (SVM), Random Forest Classifier (RFC), Extreme Gradient Boosting (XGBoost), and Decision Tree Analysis-were trained and validated using a 70:30 data split.

Results: The radiomics features demonstrated high inter- and intra-observer reproducibility. All evaluated machine learning models exhibited strong diagnostic performance, with Area Under the Curve (AUC) values exceeding 0.90. Specifically, SVM achieved an AUC of 0.922 (95% CI 0.893-0.951), RFC yielded an AUC of 0.920 (95% CI 0.881-0.963), and Decision Tree Analysis showed an AUC of 0.949 (95% CI 0.927-0.972). Notably, the XGBoost model achieved the highest diagnostic efficacy, boasting an impressive AUC of 0.987 (95% CI 0.976-0.999), coupled with a sensitivity of 89.35% and a specificity of 96.55%.

Conclusion: Our results indicate that the combination of MRI-based radiomics and machine learning algorithms, particularly XGBoost, offers a non-invasive and highly accurate method for distinguishing enchondroma from ACT.

Purpose: To evaluate the capability of large language models (LLM), specifically GPT-4 and o1, in assessing adherence to the MI-CLEAR-LLM checklist in previously published studies.

Materials and methods: A total of 159 medical research articles related to LLM applications were analyzed. Two models-GPT-4o and o1-were tested in both text-based and image-based modalities. Structured prompts incorporating reasoning strategies such as chain-of-thought and few-shot learning were used to extract information corresponding to the six core items of the MI-CLEAR-LLM checklist. Human evaluations from a prior study served as the reference standard. Each model was evaluated across three independent trials to assess consistency. Accuracy and inter-trial agreement were calculated for each checklist item.

Results: Both GPT-4o and o1 demonstrated high accuracy in extracting objective, explicitly reported items, such as LLM specifications (name, manufacturer, web access, 85.9-100%) and stochasticity parameters (63.6-95%). However, performance declined for context-dependent items, including prompt session handling (Item4, 51.5-70.7%) and test data independence (Item6, 59.6-76.8%). Text-based models generally showed superior inter-trial consistency, with GPT-4o-text achieving the highest Fleiss' kappa (κ = 0.926). In contrast, image-based models exhibited greater variability (κ = 0.402-0.772).

Conclusion: LLMs show strong potential for automating the evaluation of reporting quality in medical research, particularly for clearly structured content. However, they still face substantial challenges in extracting context-dependent or inferential information.

Background: Unresectable giant hepatocellular carcinoma (HCC) has a poor prognosis, and few effective treatment methods are available. In this study, the clinical efficacy of intensity-modulated proton therapy (IMPT) was assessed.

Methods: A retrospective review was conducted on 71 consecutive patients with giant HCC treated with IMPT at two medical centers from September 2019 to December 2022. The outcomes analyzed included liver and gastrointestinal toxicity, objective response rate (ORR), local control (LC), overall survival (OS), intrahepatic recurrence-free survival (IHRFS), and distant metastasis-free survival (DMFS).

Results: The ORR was 90.2%. The one-year and two-year LC rates were 95.0% and 86.5%, respectively. The one-year and two-year OS rates were 80.0% and 53.8%, respectively, with a median OS of 26.7 months (95% CI 20.8-32.7). At one and two years, the IHRFS rates were 47.4% and 26.0%, and the DMFS rates were 57.7% and 36.8%, respectively. Multivariate analysis indicated that ECOG performance status (p = 0.008) and complete response (CR) after IMPT were independently associated with OS (p = 0.039). Two cases (2.8%) of Grade 3 gastrointestinal toxicity were observed. The incidence of liver failure after IMPT was 12.7%, of which 6.1% were considered radiation related.

Conclusions: IMPT has potential for treating giant HCC. Despite the excellent LC and ORR, intrahepatic recurrence and distant metastasis remain significant factors affecting survival. More aggressive follow-up and additional therapeutic strategies are necessary, particularly for patients who do not achieve CR after treatment.

Purpose: To compare the accuracy and reproducibility of lesion-diameter measurements performed by three state-of-the-art LLMs with those obtained by radiologists.

Materials and methods: In this retrospective study using a public database, 83 patients with solitary colorectal-cancer liver metastases were identified. From each CT series, a radiologist extracted the single axial slice showing the maximal tumor diameter and converted it to a 512 × 512-pixel PNG image (window level 50 HU, window width 400 HU) with pixel size encoded in the filename. Three LLMs-ChatGPT-o3 (OpenAI), Gemini 2.5 Pro (Google), and Claude 4 Opus (Anthropic)-were prompted to estimate the longest lesion diameter twice, ≥ 1 week apart. Two board-certified radiologists (12 years' experience each) independently measured the same single slice images and one radiologist repeated the measurements after ≥ 1 week. Agreement was assessed with intraclass correlation coefficients (ICC); 95% confidence intervals were obtained by bootstrap resampling (5 000 iterations).

Results: Radiologist inter-observer agreement was excellent (ICC = 0.95, 95% CI 0.86-0.99); intra-observer agreement was 0.98 (95% CI 0.94-0.99). Gemini achieved good model-to-radiologist agreement (ICC = 0.81, 95% CI 0.68-0.89) and intra-model reproducibility (ICC = 0.78, 95% CI 0.65-0.87). GPT-o3 showed moderate agreement (ICC = 0.52) and poor reproducibility (ICC = 0.25); Claude showed poor agreement (ICC = 0.07) and reproducibility (ICC = 0.47).

Conclusion: LLMs do not yet match radiologists in measuring colorectal cancer liver metastasis; however, Gemini's good agreement and reproducibility highlight the rapid progress of image interpretation capability of LLMs.

Purpose: The purpose is to evaluate the utility of whole-lesion and whole-prostate gland histogram and texture analysis based on biparametric MRI (bp-MRI) for differentiating clinically significant prostate cancer (csPCa) from non-clinically significant prostate cancer (ncsPCa). We further compared the diagnostic performance of these quantitative features with PI-RADS assessment, clinical parameters, and combined models.

Materials and methods: This retrospective study enrolled 337 patients (primary cohort, 260; validation cohort, 77) with pathologically proven prostate lesions. All patients underwent preoperative prostate bp-MRI [T2-weighted imaging and apparent diffusion coefficient (ADC) maps]. Histogram and texture features were extracted from both the whole lesion and the whole-prostate gland. Diagnostic models were constructed using multivariate logistic regression, incorporating PI-RADS scores, clinical parameters, and quantitative imaging features. Their performance was evaluated using the area under the receiver operating characteristic curve (AUC) and validated on an internal cohort.

Results: Multiple histogram and texture parameters from both whole-lesion and whole-prostate analyses significantly differed between csPCa and ncsPCa groups (p < 0.05), with ADC-derived features generally outperforming T2WI-derived ones. The combined model integrating texture features, clinical parameters, and PI-RADS (Texture&Clinics&PI-RADS) demonstrated the highest diagnostic performance for both whole-lesion analysis (AUCs: 0.938 for peripheral-zone or transitional-zone (PZ + TZ), 0.894 for peripheral-zone (PZ), 0.971 for transitional-zone (TZ) lesions) and whole-prostate analysis (AUCs: 0.926 for PZ + TZ, 0.804 for PZ, 0.981 for TZ lesions) in the primary cohort. This superior performance was consistently replicated in the validation cohort. Notably, no significant difference in diagnostic efficacy was found between whole-lesion and whole-prostate analyses for TZ lesions.

Conclusion: Both whole-lesion and whole-prostate histogram and texture analysis based on bp-MRI are promising non-invasive tools for identifying csPCa. The combination of texture features, clinical parameters, and PI-RADS scores achieved the best diagnostic performance. These findings indicate that whole-lesion and whole-prostate histogram and texture analyses may improve the detection of csPCa above conventional PI-RADS assessment.

Background: Accurate prediction of metachronous liver metastasis (MLM) within the 24 months remains a clinical challenge in rectal cancer. While radiomics offers noninvasive insights into tumor characteristics, few studies have investigated multi-sequence MRI-based habitat radiomics with interpretable modeling strategies.

Methods: This retrospective study enrolled 391 patients with pathologically confirmed rectal cancer. K-means clustering was applied to pretreatment T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI) MRI to generate tumor subregions. Radiomic features were extracted from both sequences, and clinical variables were also included. Support vector machine (SVM) classifiers were used to construct radiomics, habitat, and combined models. Model performance was assessed using area under the ROC curve (AUC) and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) were employed to interpret the contribution of individual features.

Results: The habitat model demonstrated superior predictive performance compared to conventional radiomics, achieving AUCs of 0.875 in the training cohort, 0.829 in the internal validation cohort, and 0.810 in the external test cohort. The combined model, incorporating clinical variables and habitat features, achieved the highest performance in the validation cohort (AUC = 0.870) and external test cohort (AUC = 0.862). SHAP analysis revealed complementary contributions from T1WI and T2WI features, highlighting the intratumoral heterogeneity interpretability of the multi-sequence habitat approach.

Conclusion: Multi-sequence MRI-based habitat radiomics demonstrated strong performance in predicting MLM, and the integration with clinical variables further improved accuracy, providing a practical tool for individualized risk assessment and treatment planning.