Cancer Imaging最新文献

Background: This study aims to develop and validate a multi-modal machine learning model for preoperative risk stratification in papillary thyroid carcinoma (PTC), addressing limitations of current systems that rely on postoperative pathological features.

Methods: We analyzed 974 PTC patients from three medical centers in China using a multi-modal approach integrating: (1) clinical indicators, (2) immunological indices, (3) ultrasound radiomics features, and (4) CT radiomics features. Our methodology employed gradient boosting machine for feature selection and random forest for classification, with model interpretability provided through SHapley Additive exPlanations (SHAP) analysis. The model was validated on internal (n = 225) and two external cohorts (n = 51, n = 174).

Results: The final 15-feature model achieved AUCs of 0.91, 0.84, and 0.77 across validation cohorts, improving to 0.96, 0.95, and 0.89 after cohort-specific refitting. SHAP analysis revealed CT texture features, ultrasound morphological features, and immune-inflammatory markers as key predictors, with consistent patterns across validation sites despite center-specific variations. Subgroup analysis showed superior performance in tumors > 1 cm and patients without extrathyroidal extension.

Conclusion: Our multi-modal machine learning approach provides accurate preoperative risk stratification for PTC with robust cross-center applicability. This computational framework for integrating heterogeneous imaging and clinical data demonstrates the potential of multi-modal joint learning in healthcare imaging to transform clinical decision-making by enabling personalized treatment planning.

Purpose: Accurate preoperative grading of gliomas is critical for therapeutic planning and prognostic evaluation. We developed a noninvasive machine learning model leveraging whole-brain resting-state functional magnetic resonance imaging (rs-fMRI) biomarkers to discriminate high-grade (HGGs) and low-grade gliomas (LGGs) in the frontal lobe.

Methods: This retrospective study included 138 patients (78 LGGs, 60 HGGs) with left frontal gliomas. A total of 7134 features were extracted from the mean amplitude of low-frequency fluctuation (mALFF), mean fractional ALFF, mean percentage amplitude of fluctuation (mPerAF), mean regional homogeneity (mReHo) maps and resting-state functional connectivity (RSFC) matrix. Twelve predictive features were selected through Mann-Whitney U test, correlation analysis and least absolute shrinkage and selection operator method. The patients were stratified and randomized into the training and testing datasets with a 7:3 ratio. The logical regression, random forest, support vector machine (SVM) and adaptive boosting algorithms were used to establish models. The model performance was evaluated using area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity.

Results: The selected 12 features included 7 RSFC features, 4 mPerAF features, and 1 mReHo feature. Based on these features, the model was established using the SVM had an optimal performance. The accuracy in the training and testing datasets was 0.957 and 0.727, respectively. The area under the receiver operating characteristic curves was 0.972 and 0.799, respectively.

Conclusions: Our whole-brain rs-fMRI radiomics approach provides an objective tool for preoperative glioma stratification. The biological interpretability of selected features reflects distinct neuroplasticity patterns between LGGs and HGGs, advancing understanding of glioma-network interactions.

Purpose: Differentiating between benign and malignant ovarian masses remains a significant clinical challenge. Although the Ovarian-Adnexal Reporting and Data System Ultrasound Version 2022 (O-RADS US v2022) provides standardized terminology and high sensitivity, its specificity remains suboptimal, potentially leading to overdiagnosis and overtreatment. Incorporating tumor vascularity evaluation via contrast-enhanced ultrasound (CEUS) and serum tumor markers like CA125 may enhance diagnostic accuracy and help guide clinical management more effectively.

Methods: A retrospective study of 909 patients with adnexal masses undergoing ultrasound at Sichuan Cancer Hospital from May 2022 to March 2025 was conducted. O-RADS US v2022, CEUS scores, and CA125 levels were analyzed to develop a novel scoring system (OCC-US). Diagnostic performance was evaluated using ROC curves, logistic regression, and inter-observer agreement analysis. Additionally, a temporally independent validation cohort was retrospectively assembled to assess the generalizability and diagnostic accuracy of the OCC-US model.

Results: A total of 609 patients were enrolled in the development cohort between May 2022 and May 2024. ROC analysis identified O-RADS US v2022 ≥ 4, CEUS score ≥ 4, and CA125 ≥ 37.815 U/mL as independent predictors of malignancy. Based on these variables, the OCC-US scoring system was developed, assigning 2 points each for O-RADS ≥ 4 and CEUS score ≥ 4, and 1 point for CA125 ≥ 37.815 U/mL (total score range: 0-5). OCC-US achieved the highest diagnostic performance with an AUC of 0.916, outperforming OC-US (0.891), CEUS (0.877), O-RADS US v2022 (0.871), and CA125 (0.784). It significantly improved specificity (85.4% vs. 71.5%, P < 0.001) while maintaining high sensitivity (84.9%), reducing the false-positive rate from 23.1% (O-RADS US v2022) to 6.2%. OCC-US also reduced unnecessary surgical recommendations from 300 (O-RADS US v2022) to 243 (P < 0.001). Inter-observer agreement was excellent (κ = 0.840, P < 0.001), indicating high reliability. In the temporally independent external validation cohort (300 patients enrolled between June 2024 and March 2025), the OCC-US model maintained stable diagnostic performance, with an AUC of 0.867.

Conclusion: The OCC-US model enhances diagnostic specificity while maintaining high sensitivity, optimizing risk stratification and surgical decision-making. Further multi-center prospective studies are needed for broader validation.

Background: The role of ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) parameters to predict prognosis for patients with soft tissue sarcoma (STS) remains controversial.

Objectives: This meta-analysis aimed to systematically evaluate the prognostic significance of ¹⁸F-FDG PET/CT parameters in STS.

Design: This study is a systematic review and meta-analysis.

Data sources and methods: A literature search was conducted in PubMed, Embase, and the Cochrane Library for relevant studies up to January 1st, 2024. Studies exploring the association of maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) with overall survival (OS) and progression-free survival (PFS) in STS were included. Pooled hazard ratio (HR) with 95% confidence interval (CI) was calculated using random-effects models.

Results: Nineteen studies with 962 patients were included in our meta-analysis. Among these, 16 studies evaluated the correlation between the SUVmax and OS, 10 studies assessed the relationship between MTV and OS, 9 studies examined the association of TLG with OS, and 8 studies investigated the prognostic value of SUVmax in relation to PFS. The pooled HRs of SUVmax, MTV, and TLG for OS were 1.17 (95% CI: 1.07-1.27),1.87 (95% CI: 1.16-3.03), and 2.00 (95% CI: 0.99-4.01), respectively. For PFS, the pooled HR of SUVmax was1.62 (95% CI: 1.14-2.31).

Conclusion: This meta-analysis indicates that the metabolic parameter SUVmax derived from ¹⁸F-FDG PET/CT is significantly associated with poor prognosis in STS, for both OS and PFS. Additionally, MTV was significantly correlated with poor OS, whereas TLG did not show a significant relationship with prognosis in patients with STS.

Background: The impact of neck adipose tissue (NAT) on the invasiveness of hypopharyngeal squamous cell carcinoma (HPSCC) and laryngeal squamous cell carcinoma (LSCC) remains uncertain. We investigated the roles of NAT and derived - neutrophil to lymphocyte ratio (dNLR) in the aggressiveness of HPSCC and LSCC, and established an adipose- inflammation-aggressiveness axis to identify high-risk factors.

Methods: This retrospective study involved 412 patients with HPSCC or LSCC. Clinical characteristics, body mass index (BMI), NAT and dNLR were collected and calculated. Logistic regression models, restricted cubic splines (RCS) and mediation analysis were employed to evaluate the associations between NAT, dNLR and the aggressiveness of HPSCC and LSCC.

Results: The cohort included 412 patients (mean age, 63 years; 93.69% male). Lower NAT was independently associated with advanced TNM stage (adjusted Odds Ratio [OR], 0.54; p = 0.015) and tumor local invasion (adjusted OR, 0.53; p = 0.008). Higher dNLR was significantly associated with advanced TNM stage (adjusted OR, 3.26; p < 0.001), lymph node metastasis (LNM) (adjusted OR, 1.40; p = 0.021), and tumor local invasion (adjusted OR, 2.29; p < 0.001). NAT showed a modest negative correlation with dNLR (R = -0.138, p = 0.005). Mediation analysis indicated that dNLR partially mediated the relationship between NAT and tumor aggressiveness.

Conclusions: Reduced NAT is associated with increased tumor aggressiveness in HPSCC and LSCC, and this relationship may be partially mediated by elevated dNLR. The association appeared more pronounced in male patients. These findings suggest that local adiposity and inflammation may play a role in tumor behavior and warrant further investigation in future studies.

Objectives: This study aims to assess the predictive capability of PSMA-PET imaging for disease outcomes in primary prostate cancer post-radical prostatectomy. In addition to conventional lesion uptake measures, the evaluation includes the distance of lesion to the prostate to enhance risk stratification and outcome prediction.

Methods: A cohort of 190 men diagnosed with primary prostate cancer and undergoing prostatectomy were initially screened, resulting in 103 patients meeting the selection criteria. Imaging parameters, including lesion SUVmax, primary metabolic tumor volume (PMTV), maximum distance from the lesion to the prostate (Dmax), and total distances from the lesion to the prostate (Dtotal), were extracted from 68Ga-PSMA-11 PET images. Findings were dichotomized based on primary lesion uptake, the tumor volume size, Dmax distance, and the presence of metastatic disease. Postoperative biochemical recurrence-free survival (BCRFS) was analyzed using Kaplan-Meier survival plots and Log-rank tests. Furthermore, univariate and multivariate Cox regression analyses were performed to evaluate the association of PET parameters with survival outcomes.

Results: Clinical and histopathological characteristics were summarized, including age, weight, height, metastasis status, baseline PSA, biopsy Gleason score, pt stage, margin status, and lymph node status. After a median follow-up of 20 months, 66 events occurred, with the estimated 3-year BCRFS being 46%. Increased PSMA intensity (SUVmax > 17.06) was associated with less favorable BCRFS (log-rank p = 0.017). Increased primary metabolic tumor volume (PMTV > 41.59 cm³) was also linked to less favorable BCRFS (log-rank p = 0.003). Dmax and Dtotal greater than 9.69 cm and 11.95 cm were identified as negative prognostic factors for BCRFS (log-rank p < 0.001 and p = 0.002, respectively). Based on PMTV and Dmax, patients were stratified into low-, intermediate-, and high-risk groups, with 3-year BCRFS rates of 57%, 31%, and 8%, respectively. Univariate Cox regression analysis revealed significant associations between BCRFS and factors such as baseline PSA (HR: 1.69, 95% CI 1.02-2.79, p = 0.042), SUVmax (HR: 1.56, 95% CI 1.04-1.91, p = 0.018), PMTV (HR: 2.05, 95% CI 1.26-3.34, p = 0.004), Dmax (HR: 2.24, 95% CI 1.37-3.65, p = 0.001), and Dtotal (HR: 2.11, 95% CI 1.29-3.45, p = 0.003). Multivariable Cox regression analysis identified the best model with PMTV (HR: 2.57, p = 0.004) and Dmax (HR: 1.98, p = 0.009) as independent predictors for biochemical recurrence (C-index = 0.68).

Conclusion: The lesion distance to prostate was defined and assessed in conjunction with conventional PET parameters to facilitate preoperative risk stratification in primary prostate cancer following radical prostatectomy. The findings contribute to improved outcome prediction and emphasize the potential of PSMA-PET imaging in enh

Background: While the TNM staging system provides valuable insights into the extent of disease, predicting postoperative progression in early-stage non-small cell lung cancer (NSCLC) remains a significant challenge. An effective bioimaging prognostic marker for early-stage NSCLC, powered by artificial intelligence, could greatly assist clinicians in making informed treatment decisions.

Methods: A total of 926 patients from four centers (A, B, C, and D) with histologically confirmed stage I or II solid non-small cell lung cancer (NSCLC) who underwent surgical resection were retrospectively reviewed. In this study, we propose a robust federated learning model (RFed) designed to predict the risk of postoperative progression in early-stage NSCLC patients. The diagnostic efficiency of the RFed model was evaluated using the area under the curve (AUC) and Decision Curve Analysis (DCA). Additionally, the model's performance was further validated through Kaplan-Meier survival analysis, with statistical significance assessed using the log-rank test. Finally, the robustness, generalizability, and interpretability of the RFed model were comprehensively evaluated to confirm its clinical applicability.

Results: Experimental results demonstrated the superior performance of the RFed model. Specifically, RFed achieved AUC values of 0.936, 0.861, 0.925, and 0.970 on the test sets from the four centers. DCA further revealed that RFed provided a greater net benefit compared to the clinical model across a threshold probability range of 0.02 to 0.99. Moreover, Kaplan-Meier curves showed improved discrimination between high-risk and low-risk groups when compared to other models, highlighting its enhanced predictive capability.

Conclusions: The RFed model demonstrates significant effectiveness in predicting the risk of postoperative progression in early-stage NSCLC patients. Its clinical application value lies in its potential to enhance stratified management and support the development of precise treatment strategies for this patient population.

Background: The tumour sink effect is a phenomenon whereby the sequestration of a radiopharmaceutical in cancer lesions leads to decreased activity concentration in the blood stream and organs. The aim of this sub-analysis of the prospective 3TMPO study (NCT04000776) was to investigate the tumour sink effect on prostate-specific membrane antigen (PSMA) PET imaging in a population of patients with metastatic castration-resistant prostate cancer (mCRPC).

Methods: Ninety-seven participants underwent ⁶⁸Ga-PSMA-617 PET/CT imaging. The activity concentration in the kidney, parotid, spleen, liver and blood was expressed as a percentage of injected activity per cubic centimetre (%IA/cm³). The total tumour volume was delineated, and the total lesion fraction (TLF), i.e., the percentage of injected activity sequestered in the tumour, was computed. Participants were stratified into three tumour burden groups: small (TLF < 10%), moderate (10% ≤ TLF < 25%), and large (TLF ≥ 25%). Weight, lean body weight, body surface area, and estimated glomerular filtration rate (eGFR) were investigated as additional factors affecting biodistribution.

Results: The TLF ranged from 0.0 to 43.5%. For all healthy tissues, the %IA/cm³ was negatively correlated with TLF (r ranging - 0.33 to - 0.46; P < 0.001). Patients with a large TLF had significantly lower uptake in all organs when compared to those with a small TLF (P < 0.05). Body habitus indices and/or eGFR were negatively correlated with the %IA/cm³ of the parotid, liver and blood (r ranging - 0.23 to - 0.33; P < 0.05). Combining predictive variables, the term [BSA / (1-TLF)] tended to yield the strongest negative correlations with healthy tissues %IA/cm³ (r ranging - 0.33 to - 0.63; P < 0.001).

Conclusion: The tumour sink effect was observed in a cohort of mCRPC patients scanned with ⁶⁸Ga-PSMA-617. This finding strongly suggests that patients with a large TLF are likely to receive lower absorbed doses to organs at risk - i.e., be undertreated from a dosimetry perspective - following a fixed-activity regime of ¹⁷⁷Lu-PSMA-617 radiopharmaceutical therapy, as commonly practiced. Individual factors such as body habitus and renal function further impact the biodistribution of PSMA radiopharmaceuticals.

Trial registration: NCT04000776, registered on 2019-06-27.

Objective: Accurate preoperative imaging localization is paramount to the success of targeted parathyroidectomy for primary hyperparathyroidism (PHPT). Four-dimensional (4D) CT is a promising method for preoperative localization of the parathyroid, but studies on the performance of 4D CT and technetium 99 m-sestamibi SPECT/CT for the diagnosis of diseases of the parathyroid are limited.

Materials and methods: To compare the diagnostic performance of sestamibi SPECT/CT and 4D-CT for preoperative localization in patients with PHPT in a single-institution from August 2017 to May 2024.

Results: Two hundred forty-two patients with PHPT (166 females; 52.5 years ± 13.4 [SD]) were evaluated. Among the 242 patients, 233 patients (96.3%) had single-gland disease, and 9 patients (3.7%) had multigland disease. Similar diagnostic performance was observed for sestamibi SPECT/CT and 4D-CT ([receiver operating characteristic ROC], 0.90 [95% CI: 0.87, 0.92] and 0.88 [95% CI: 0.85, 0.90], respectively; p = 0.11). Compared with 4D-CT, combined-modality sensitive reading and sestamibi SPECT/CT had the highest ROC, and, although there was no significant difference between the two (ROC, 0.91; 95% CI: 0.89, 0.93; p = 0.14), they significantly differed from 4D-CT (p = 0.0006). Sestamibi SPECT/CT showed an accuracy of 92% (95% CI: 90%, 94%), similar to 4D-CT (91%; 95% CI: 89%, 92%), combined-modality sensitive reading (91%; 95% CI: 89%, 93%) and combined-modality specificity reading (92%; 95% CI: 90%, 94%).

Conclusion: Sestamibi SPECT/CT has high accuracy in preoperative localization in patients with PHPT. Compared with sestamibi SPECT/CT alone, 4D-CT and combined-modality reading did not improve diagnostic performance.

Purpose: This study assessed the effectiveness and prognostic factors of CT-guided ¹²⁵I seed implantation for primary hepatocellular carcinoma (HCC).

Methods: A retrospective analysis of 71 patients (57 males, 14 females, median age 64) treated at three Chinese hospitals from 2018 to 2024 was conducted. The main outcomes were local progression-free survival (LPFS) and overall survival (OS). Treatment involved a 16-slice Spiral CT and Radioactive Particle Treatment Planning System (TPS), with seeds of 18.5-29.6 MBq implanted via freehand puncture. Efficacy was evaluated using the modified Response Evaluation Criteria in Solid Tumors (mRECIST) at three months, with follow-ups every three months for three years, then biannually until December 2024. Data analysis utilized SPSS 22.0, Kaplan-Meier, and Cox models.

Results: With a median follow-up of 37 months, the complete response (CR) rate was 57.7%, partial response (PR) 31.0%, stable disease (SD) 5.6%, and progressive disease (PD) 5.6%. Local control was 94.3%. LPFS rates at 1, 3, and 5 years were 74.6%, 29.5%, and 1.4% (median LPFS 22 months), while overall survival (OS) rates were 88.7%, 47.8%, and 12.6% (median OS 35 months). CR was a key protective factor for LPFS and OS. Significant factors included the Barcelona Clinic Liver Cancer (BCLC) stage C, intrahepatic progression, and extrahepatic metastasis. Postoperative complications occurred in 35.2% of patients, with no severe cases.

Conclusion: CT-guided ¹²⁵I seed implantation is effective and safe for primary HCC, with CR being crucial for survival. Large-scale studies are needed to confirm these results.