Academic Radiology最新文献

Rationale and objectives: The prognostic role of computed tomography (CT)-defined skeletal muscle features in COVID-19 is still under investigation. The aim of the present study was to evaluate the prognostic role of CT-defined skeletal muscle area and density in patients with COVID-19 in a multicenter setting.

Materials and methods: This retrospective study is a part of the German multicenter project RACOON (Radiological Cooperative Network of the COVID-19 pandemic). The acquired sample included 1379 patients, 389 (28.2%) women and 990 (71.8%) men. In each case, chest CT was analyzed and pectoralis muscle area and density were calculated. Data were analyzed by means of descriptive statistics. Group differences were calculated using the Mann-Whitney-U test and Fisher's exact test. Univariable and multivariable logistic regression analyses were performed.

Results: The 30-day mortality was 17.9%. Using median values as thresholds, low pectoralis muscle density (LPMD) was a strong and independent predictor of 30-day mortality, HR=2.97, 95%-CI: 1.52-5.80, p=0.001. Also in male patients, LPMD predicted independently 30-day mortality, HR=2.96, 95%-CI: 1.42-6.18, p=0.004. In female patients, the analyzed pectoralis muscle parameters did not predict 30-day mortality. For patients under 60 years of age, LPMD was strongly associated with 30-day mortality, HR=2.72, 95%-CI: 1.17;6.30, p=0.019. For patients over 60 years of age, pectoralis muscle parameters could not predict 30-day mortality.

Conclusion: In male patients with COVID-19, low pectoralis muscle density is strongly associated with 30-day mortality and can be used for risk stratification. In female patients with COVID-19, pectoralis muscle parameters cannot predict 30-day mortality.

Rationale and objectives: To comprehensively assess the feasibility of low-dose computed tomography (LDCT) using deep learning image reconstruction (DLIR) for evaluating pulmonary subsolid nodules, which are challenging due to their susceptibility to noise.

Materials and methods: Patients undergoing both standard-dose CT (SDCT) and LDCT between March and June 2023 were prospectively enrolled. LDCT images were reconstructed with high-strength DLIR (DLIR-H), medium-strength DLIR (DLIR-M), adaptive statistical iterative reconstruction-V level 50% (ASIR-V-50%), and filtered back projection (FBP); SDCT with FBP as the reference standard. Objective assessment, including image noise, contrast-to-noise ratio (CNR), and signal-to-noise ratio (SNR), and subjective assessment using five-point scales by five radiologists were performed. Detection and false-positive rate of subsolid nodules, and morphologic features of nodules were recorded.

Results: 102 patients (mean age, 57.0 ± 12.3 years) with 358 subsolid nodules in SDCT were enrolled. The mean effective dose of SDCT and LDCT were 5.37 ± 0.80mSv and 0.86 ± 0.14mSv, respectively (P < 0.001). DLIR-H showed the lowest noise, highest CNRs, SNRs, and subjective scores among LDCT groups (all P < 0.001), almost approaching comparability with SDCT. The detection rates for DLIR-H, DLIR-M, ASIR-V-50%, and FBP were 76.5%, 76.3%, 83.8%, and 72.1%, respectively (P < 0.001), with false-positive rate of 2.5%, 2.2%, 8.3%, and 1.1%, respectively (P < 0.001). DLIR-H showed the highest detection rates for morphologic features (79.4%-95.2%) compared to DLIR-M (74.6%-88.9%), ASIR-V-50% (72.0%-88.4%), and FBP (66.1%-84.1%) (all P ≤ 0.001).

Conclusion: Sub-milliSievert LDCT with DLIR-H offers substantial dose reduction without compromising image quality. It is promising for evaluating subsolid nodules with a high detection rate and better identification of morphologic features.

Rationale and objectives: Angioleiomyomas are benign perivascular tumors that originate from the tunica media of blood vessels. While frequently described in the head, neck, and uterus, angioleiomyomas can manifest in various regions throughout the body. The purpose of this study was to review the history and imaging features of angioleiomyomas of the trunk and extremities.

Materials and methods: Patients with pathologically proven angioleiomyomas at our institution were retrospectively identified. Clinical information was obtained by chart review. Any available imaging of the tumor was reviewed.

Results: This study includes 191 patients with angioleiomyoma of the trunk or extremities, 87 with imaging of the tumor. Mean age at presentation was 55.5 years and 59.7% of patients were female. The tumor was painful in 88.9% of patients. Most lesions were in the lower extremity (79.1%), followed by the upper extremity (17.8%) and trunk (3.1%). A nonspecific soft tissue mass was visible radiographically in 27.4% of cases, with calcifications in 1.8%. On ultrasound, the tumor was always hypoechoic, with internal vascularity in 93.8%. Most tumors were T1 isointense and T2 hyperintense relative to skeletal muscle (92.9%) and enhanced (95.8%). CT showed a soft tissue density mass in all cases. On cross-sectional imaging, the mass was directly adjacent to a blood vessel in 83.1% of cases.

Discussion: Key imaging features of angioleiomyomas include a soft tissue mass with adjacent blood vessel on cross-sectional imaging. Ultrasound shows a hypoechoic mass with internal vascularity. They are typically T1 isointense, T2 hyperintense enhancing masses which may have a dark reticular sign and/or hypointense peripheral rim. Recognizing these features may help include angioleiomyoma in the differential diagnosis.

Rationale and objectives: High-grade patterns, visceral pleural invasion, lymphovascular invasion, spread through air spaces, and lymph node metastasis are high-risk factors and associated with poor prognosis in lung adenocarcinomas (LUADs). This study aimed to construct and validate a radiomic model and a radiographic model derived from low-dose CT (LDCT) for predicting high-risk LUADs in solid and part-solid nodules.

Materials and methods: This study retrospectively enrolled 658 pathologically confirmed LUADs from July 2018 to December 2022 from four centers, which were divided into training set (n=411), internal validation set (n=139), and external validation set (n=108). Radiomic features and radiographic features including maximal diameter, consolidation/tumor ratio (CTR), and semantic features, were obtained to construct a radiomic model and a radiographic model through multivariable logistic regression. Area under receiver operating characteristic curve (AUC) was utilized to assess the diagnostic performance of the models.

Results: Three radiomic features (GLCM_Correlation, GLSZM_SmallAreaEmphasis, and GLDM_LargeDependenceHighGrayLevelEmphasis) and four radiographic features (maximal diameter, CTR, spiculation, and pleural indentation) were selected to build models. The radiomic model yielded AUCs of 0.916 in the internal validation set and 0.938 in the external validation set, which were significantly higher than the AUCs of the radiographic model (0.916 vs. 0.868, P=0.014 and 0.938 vs. 0.880, P=0.002).

Conclusion: Our LDCT-based radiomic model enabled non-invasive identification of high-risk LUADs in solid and part-solid nodules with good diagnostic performance and might assist in case-specific decision-making in lung cancer screening.

Purpose: This study aimed to develop and evaluate a machine learning model combining clinical, radiomics, and deep learning features derived from PET/CT imaging to predict lymph node metastasis (LNM) in patients with non-small cell lung cancer (NSCLC). The model's interpretability was enhanced using Shapley additive explanations (SHAP).

Methods: A total of 248 NSCLC patients who underwent preoperative PET/CT scans were included and divided into training, test, and external validation sets. Radiomics features were extracted from segmented tumor regions on PET/CT images, and deep learning features were generated using the ResNet50 architecture. Feature selection was performed using minimum-redundancy maximum-relevance (mRMR), and the least absolute shrinkage and selection operator (LASSO) algorithm. Four models-clinical, radiomics, deep learning radiomics (DL_radiomics), and combined model-were constructed using the XGBoost algorithm and evaluated based on diagnostic performance metrics, including area under the receiver operating characteristic curve (AUC), accuracy, F1 score, sensitivity, and specificity. Shapley Additive exPlanations (SHAP) was used for model interpretability.

Results: The combined model achieved the highest AUC in the test set (AUC=0.853), outperforming the clinical (AUC=0.758), radiomics (AUC=0.831), and DL_radiomics (AUC=0.834) models. Decision curve analysis (DCA) demonstrated that the combined model offered greater clinical net benefits. SHAP was used for global interpretation, and the summary plot indicated that the features ct_original_glrlm_LongRunHighGrayLevelEmphasis, and pet_gradient_glcm_lmc1 were the most important for the model's predictions.

Conclusion: The combined model, combining clinical, radiomics, and deep learning features from PET/CT, significantly improved the accuracy of LNM prediction in NSCLC patients. SHAP-based interpretability provided valuable insights into the model's decision-making process, enhancing its potential clinical application for preoperative decision-making in NSCLC.

Rationale and objectives: The purpose of this study was to explore intra-individual differences in pericoronary adipose tissue (PCAT) fat attenuation index (FAI) between photon-counting detector (PCD)- and energy-integrating detector (EID)-CT.

Material and methods: Patients were prospectively enrolled for a PCD-CT research scan within 30 days of EID-CT. Both acquisitions were reconstructed using a Qr36 kernel at 0.6 mm slice thickness (EID and PCD-down-sampled [DS]) and at 0.2 mm ultra-high resolution (UHR) for the PCD-CT. Iterative reconstruction was turned "off" (filter back projection used as alternative reconstruction method) or set to a recommended level in current literature. Coronary PCAT FAI was measured automatically using established thresholds of -190 to -30 HU at a set distance and radius. Statistical testing was performed using repeated-measures ANOVA and Bonferroni's multiple comparison tests (p significance was determined to be <0.003).

Results: In total, 40 patients (mean age 68±8 years, 32 males [80%]) were included for analysis. Absolute FAI measurements differed significantly for all vessels between all reconstructions in the ANOVA comparison (all p<.001). The FAI decreased going from EID-CT to PCD-DS, to PCD-UHR with iterative reconstruction turned off (e.g. right coronary artery: EID-CT: -76.5±8.9 vs -80.9±7.0 vs -88.3±6.7 HU, respectively; p < 0.001). The mean FAI of datasets using iterative reconstruction did not demonstrate significant differences on multiple comparisons (e.g. left circumflex artery: EID: -65.7±8.5; PCD-DS: -66.0±7.4; PCD-UHR: -67.8±7.0 HU, respectively; p>0.06).

Conclusion: Intra-individual absolute PCAT FAI measurements differ significantly between EID- and PCD-CT when controlling for reconstruction kernel and slice thickness. However, the use of iterative reconstruction minimizes most differences in FAI, enabling inter-scanner comparability.

Rationale and objectives: We aimed to compare the capabilities of two leading large language models (LLMs), GPT-4 and Gemini, in analyzing serial radiology reports, to highlight oncological issues that require further clinical attention.

Materials and methods: This study included 205 patients, each with two consecutive radiological reports. We designed a prompt comprising a three-step task to analyze report findings using LLMs. To establish a ground truth, two radiologists reached a consensus on a six-level categorization, comprising tumor findings (categorized as improved, stable, or aggravated), "benign", "no tumor description," and "other malignancy." The performance of GPT-4 and Gemini was then compared based on their ability to match corresponding findings between two radiological reports and accurately reflect these categories.

Results: In terms of accuracy in matching findings between serial reports, the proportion of correctly matched findings was significantly higher for GPT-4 (96.2%) than for Gemini (91.7%) (P < 0.01). For oncological issue identification, the precision for tumor-related finding determinations, recall, and F1-scores were 0.68 and 0.63 (P = 0.006), 0.91 and 0.80 (P < 0.001), and 0.78 and 0.70 for GPT-4 and Gemini, respectively. GPT-4 was more accurate than Gemini in determining the correct tumor status for tumor-related findings (P < 0.001).

Conclusion: This study demonstrated the potential of LLM-assisted analysis of serial radiology reports in enhancing oncological surveillance, using a carefully engineered prompt. GPT-4 showed superior performance compared to Gemini in matching corresponding findings, identifying tumor-related findings, and accurately determining tumor status.

Rationale and objectives: The correlation between collateral circulation and futile recanalization (FR) is still controversial, and few studies have explored the influence of comprehensive cerebral collateral circulation on FR after endovascular stroke treatment. Therefore, based on cerebral collateral recycle (CCR) status, we aimed to establish an effective scoring system to identify the probability of FR.

Methods: This was a multicenter retrospective cohort study. FR was defined as a 90-day modified Rankin Scale (mRS) score of 3-6, despite having successful recanalization (modified Thrombolysis in Cerebral Infarction score of 2b-3). The discrimination and calibration of this score were assessed using the area under the receiver operator characteristic curve, calibration curve, and decision curve analysis.

Results: Out of 860 patients receiving endovascular stroke treatment, 478 were enrolled in this study after strict screening. In multivariate regression analysis, the CCR status (poor CCR, adjusted OR[aOR] 9.99, 95%CI 5.11 to 17.06, P < 0.001; moderate CCR, aOR 2.94, 95%CI 1.71 -5.06, P < 0.001), age ≥ 80 years (aOR 3.77, P < 0.001), baseline NIHSS ≥ 15 (aOR 1.81, P = 0.018), baseline ASPECTS ≤ 6 (aOR 1.95, P = 0.006), the time from stroke onset to revascularization (OTR) ≥ 600 min (aOR 2.02, P = 0.007) and any intracranial hemorrhage within 48 h (aOR 3.54, P < 0.001) were significantly associated with FR. These factors make up the CCR-hemorrhage-age-NIHSS-OTR-ASPECTS (CHANOA) score. The CHANOA score demonstrated good discrimination and calibration in this cohort, as well as the fivefold cross validation.

Conclusion: The CHANOA score reliably predicted FR in patients with endovascular stroke treatment, based on comprehensive cerebral collateral and clinical features.