Insights into Imaging最新文献

Objectives: To assess a deep learning (DL) model using portal-venous phase CT for discriminating colorectal cancer liver metastasis (CRLMs) and hemangiomas (HMs).

Materials and methods: Colorectal cancer (CRC) patients diagnosed with CRLMs or HMs at two medical centers from January 2018 and April 2024 were retrospectively included. Lesions were automatically segmented using TotalSegmentator. DL models, DenseNet-201 and ResNet-152, were trained to classify CRLMs and HMs. Their performance, measured by AUC, was evaluated on validation and test sets. Subgroup analyses were conducted for lesions ≤ 10 mm (subcentimeter) and 10-30 mm. Radiologists' diagnostic performance with and without DL assistance was compared using a multi-reader multi-case analysis.

Results: 534 CRLMs (134 CRC-patients; median, 60 years) and 262 HMs (154 CRC-patients; median, 62 years) were divided into the training, validation and test set. The Dice coefficients of TotalSegmentor for automatically segmenting subcentimeter and 10-30 mm lesions were 0.692 ± 0.099 and 0.861 ± 0.033, respectively (p < 0.01). ResNet-152 model achieved AUCs of 0.875 (95% CI: 0.838-0.912), 0.858 (95% CI: 0.781-0.935), 0.776 (95% CI: 0.703-0.848) for classifying CRLMs and HMs on the training, validation, and test sets, respectively. The AUCs for distinguishing between 10-30 mm CRLMs and HMs improved from 0.851 (95% CI: 0.821-0.880) to 0.879 (95% CI: 0.853-0.906) with DL assistance compared to without (p = 0.015). For subcentimeter CRLMs and HMs, the AUCs for the radiologists and the DL-assisted diagnosis were 0.742 (95% CI: 0.669-0.814) and 0.763 (95% CI: 0.681-0.845), respectively (p = 0.558).

Conclusion: DL can assist radiologists in distinguishing 10-30 mm CRLMs from HMs in CRC patients. The value of DL-assisted diagnosis is limited for subcentimetre CRLMs and HMs.

Critical relevance statement: Dynamic detection of hypoenhancing liver lesions in patients with CRC is exceptionally challenging. The DL tool we have developed can assist in evaluating CRLMs and HMs.

Key points: TotalSegmentator can perform automatic segmentation of CRLMs and HMs, but demonstrates poorer segmentation consistency for subcentimeter lesions. This DL model assists radiologists in distinguishing 10-30 mm CRLMs from HMs in CRC patients. Subcentimeter CRLMs and HMs can require further MRI scanning.

Objectives: To evaluate the feasibility and safety (primary endpoints), and performance (secondary endpoint) of a new artificial intelligence (AI) software for detecting clinically significant prostate cancer (csPCa) on biparametric MRI (bpMRI) compared to an expert radiologist.

Materials and methods: In this prospective study at St. Olavs Hospital, Norway (December 2023-October 2024), 89 consecutive biopsy-naïve men underwent bpMRI for suspected PCa. Scans were interpreted by a radiologist using PI-RADS v2.1 and a radiomics-based AI software. Biopsies were obtained from all radiologist- and/or AI-identified lesions. csPCa was defined as ISUP ≥ 2. Feasibility was defined by a < 10% software-failure rate, and safety by the absence of serious adverse device effects (SADEs). Performance was evaluated with ROC, free-response ROC, and precision-recall curves.

Results: Among 89 patients eligible for primary endpoints evaluation, the software demonstrated feasibility (7% failure rate) and safety (no SADEs). Among 76 eligible for secondary endpoint evaluation (median age 68 years [IQR: 63-73]), csPCa was found in 51% (39/76). Patient-level, software achieved an area under the ROC curve [95% CI] of 0.90 [0.83, 0.96] versus 0.86 [0.76, 0.93] (p = 0.25). At a retrospectively optimized threshold matching the radiologist's patient-level sensitivity at PI-RADS 3 (0.92), software achieved specificity of 0.68 [0.57, 0.78] versus 0.57 [0.46, 0.68] (p = 0.29). Lesion-level, software achieved higher average precision (0.61 [0.52, 0.71] vs. 0.56 [0.46, 0.67]) and lower average false-positive per patient (0.33 [0.22, 0.43] vs. 0.41 [0.30, 0.52]) at the optimized threshold.

Conclusion: The software was feasible and safe, and diagnostic performance showed potential to reduce unnecessary biopsies.

Critical relevance statement: This clinically validated artificial intelligence software enables feasible and safe detection of clinically significant prostate cancer on biparametric MRI, with demonstrated potential to reduce unnecessary biopsies and improve diagnostic accuracy, indicating potential for integration into clinical prostate cancer care.

Key points: A fully automated radiomics software for clinically significant prostate cancer detection on biparametric MRI was prospectively clinically validated. The software demonstrated feasibility and safety, with potential to reduce unnecessary biopsies and improve diagnostic accuracy. The investigated radiomics software has the potential for integration into clinical prostate cancer care.

Objective: Deep learning signatures (DLS) extracted from CT images can noninvasively reflect tumor heterogeneity and have shown promise in prognostic modeling for esophageal squamous cell carcinoma (ESCC). To develop and validate a CT-based DL model combined with nutritional biomarkers to predict 3-year overall survival (OS) in ESCC, and to investigate transcriptomic differences between DLS-based risk groups.

Materials and methods: This retrospective multicenter study included 662 postoperative ESCC patients from three hospitals and 16 additional patients from The Cancer Genome Atlas (TCGA). DL features extraction from CT images based on the Crossformer architecture. Skeletal muscle index was measured at the L3 vertebra to assess low skeletal muscle mass (LSMM). Cox regression was used to build clinical, DL, and combined models. Model performance was evaluated using the concordance index (C-index). Transcriptomic analysis of the TCGA cohort was performed to identify metabolic pathway differences between DLS-based risk groups.

Results: The DL model achieved a C-index of 0.743 (95% CI: 0.683-0.803) in the internal validation cohort and 0.692 (95% CI: 0.576-0.809) in the external cohort. Pathological T and N stages, Neuroaggression, Vascular invasion, and LSMM were identified as independent clinical predictors. The combined model achieved a C-index of 0.753 (95% CI: 0.697-0.808) internally and 0.725 (95% CI: 0.613-0.838) externally. DLS-based risk stratification revealed significant differences in metabolic activity between groups, supporting its biological relevance.

Conclusion: The combined model enables preoperative OS prediction in ESCC. DLS-based stratification reflects transcriptomic metabolic heterogeneity and enhances the biological interpretability of imaging features.

Critical relevance statement: This study developed a CT-based DLS and combined it with nutritional markers for prognostic modeling in ESCC. Transcriptomic analysis of DLS-based groups revealed metabolic heterogeneity, enhancing the biological interpretability of the DL model.

Key points: A combined DLS and nutritional model enables individualized preoperative survival prediction in ESCC. DLS-based risk groups defined by the DLS exhibited transcriptomic differences in key metabolic pathways, revealing biological underpinnings of imaging-based phenotypes. Attention map visualization revealed consistent spatial focus on morphologically distinct tumor regions, enhancing the interpretability of deep learning predictions.

Objectives: Waiting for postoperative pathologic confirmation of visceral pleural invasion (VPI) may delay treatment decisions. This study aimed to develop a contrast-enhanced CT-based radiomics model for preoperative prediction of VPI in early-stage non-small cell lung cancer (NSCLC).

Materials and methods: We retrospectively enrolled 523 surgically resected NSCLC patients (195 with VPI, 328 without VPI) with clinically staged IA based on preoperative imaging between December 2019 and June 2022. Patients were randomly divided into training, validation, and testing sets at a ratio of 5:2:3. For each patient, 13 CT features were recorded, including the types I-V tumor relationships to the pleura. Regions of interest (ROIs) were segmented semi-automatically using deep learning. Least absolute shrinkage and selection operator (LASSO) regression was applied to select key radiomics features. Three models were developed: a CT-feature model, a radiomics model, and a combined model. The performance and clinical utility of these models were evaluated using the area under the curve (AUC) and decision curve analysis.

Results: The tumor relationship to the pleura, density, maximum diameter, and spiculation were selected to construct the CT-feature model. A total of 10 optimal features formed the radiomics model. The radiomics model achieved an AUC of 0.812 in the testing set, outperforming the CT-feature model (0.714). Furthermore, the combined model showed a slightly higher AUC (0.825) compared to the radiomics model.

Conclusions: The radiomics model demonstrated satisfactory performance for predicting VPI in early-stage NSCLC, outperforming the CT-feature model. The integration of radiomics and CT features may provide enhanced predictive value.

Critical relevance statement: This study constructed a contrast-enhanced CT-based radiomics model with promising performance for the preoperative prediction of VPI, which aims to guide treatment planning for early-stage NSCLC.

Key points: VPI affects the tumor-node-metastasis (TNM) staging of tumors and subsequent treatment strategies. The radiomics model outperformed the CT-feature model in predicting VPI. The contrast-enhanced CT-based radiomics model may be valuable for optimizing clinical decision-making.

Objectives: Androgen deprivation therapy (ADT) is essential for treating prostate cancer (PCa) but is limited by tumor heterogeneity. This study develops a non-invasive multiparametric Magnetic Resonance Imaging (mpMRI) radiomics framework to predict ADT response and improve risk stratification.

Materials and methods: A cohort of 550 ADT-treated PCa patients from three centers was analyzed. Patients were randomly divided into training (n = 270) and internal validation (n = 115) cohorts. An external test cohort (n = 165) from Centers 2 and 3 was used for generalizability. Radiomics models based on T2-weighted and diffusion-weighted imaging (DWI), habitat radiomics, and a 3D Vision Transformer (ViT) deep learning model were developed. Ensemble integration of these models was performed, with SHapley Additive exPlanations (SHAP) used for interpretability. Predictive performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC).

Results: Habitat radiomics outperformed conventional radiomics in Gleason score stratification. For predicting ADT treatment efficacy, the radiomics model achieved AUCs of 0.969 (training), 0.767 (internal validation), and 0.771 (test). The habitat model showed AUCs of 0.987, 0.849, and 0.820, while the ViT model achieved AUCs of 0.831, 0.805, and 0.796. The ensemble model reached the highest AUC of 0.886. SHAP analysis shows that the ViT model contributes most to the combined model, followed by the habitat model, with the radiomics model contributing the least.

Conclusion: mpMRI-based habitat radiomics enables precise risk stratification in PCa. Integrated with conventional radiomics and deep learning, it forms a robust framework for predicting ADT response and guiding personalized treatment.

Critical relevance statement: This study demonstrates that integrating habitat radiomics with deep learning improves the prediction of androgen deprivation therapy response in PCa, advancing personalized radiological decision-making through interpretable multi-model analysis of tumor microenvironment heterogeneity.

Key points: Multi-model integration of habitat radiomics and 3D Vision Transformer achieves superior prediction for ADT response compared to conventional methods. Habitat radiomics outperforms traditional radiomics in Gleason score stratification. SHAP analysis provides clinical interpretability, identifying key model linked to ADT outcomes for actionable insights.

Objectives: To propose an easy-to-use binary scoring system for background signal intensity changes in prostate MRI that may affect diagnostic image interpretation and to evaluate its impact on cancer detection.

Materials and methods: This retrospective single-center study included 200 patients. Four readers independently assigned background scores of A or B according to the proposed scoring system and assessed the presence or absence of cancer. Light's kappa was used to evaluate inter-reader agreement on the score and on the presence of clinically significant prostate cancer in dependence of the score. Sensitivity and specificity in detecting clinically significant cancer were assessed relative to histology as the gold standard.

Results: Due to suboptimal image quality according to the PI-QUAL score, 45 patients were secondarily excluded. Inter-reader agreement on the score was substantial (kappa = 0.62, 95% CI = 0.54-0.71). Inter-reader agreement on the presence of cancer was higher for a background score A (kappa = 0.49, 95% CI = 0.38-0.61) than B (kappa = 0.34, 95% CI = 0.20-0.51). Sensitivity in detecting cancer was high regardless of the background score (86.61% and 89.42% for scores A and B), while specificity decreased markedly in readers with little experience (53.47% and 43.75% for scores A and B), potentially increasing false positives.

Conclusion: After further validation, the easy-to-use binary background score could enable routine evaluation of normal changes in the peripheral zone, identifying cases with increased false-positive risk among inexperienced readers.

Critical relevance statement: The easy-to-use binary background score for daily clinical routine allows the communication of potential diagnostic uncertainties in mpMRI image interpretation of the prostate that arise due to normal changes in the peripheral zone, especially for less experienced readers.

Key points: An easy-to-use binary scoring system for addressing background signal intensity changes in the prostate is proposed for MRI interpretation. Inter-reader agreement of the score was substantial, and agreement between readers regarding the presence or absence of cancer was higher for a background score of A than B. The background score could be used to communicate a potential diagnostic uncertainty related to the normal change in the peripheral zone, particularly for less experienced readers.

Objectives: In this retrospective study, we aimed to assess the predictive value of the Carotid Plaque-RADS (Reporting and Data System) for coronary functional stenosis in candidates for carotid revascularization, using high-resolution magnetic resonance imaging (HR-MRI) coupled with computed tomography-derived fractional flow reserve (CT-FFR).

Materials and methods: A retrospective analysis was performed on data of 101 patients with carotid atherosclerosis who underwent HR-MRI for Carotid Plaque evaluation, and CT-FFR for coronary assessment was conducted. Patients were divided into two groups based on a CT-FFR threshold of ≤ 0.80. Logistic regression, correlation analyses, and receiver operating characteristic curve analyses were used to identify predictors of coronary functional stenosis.

Results: In the functional stenosis group (n = 76), both plaque volume and Carotid Plaque-RADS categories had higher values than those observed in the non-functional group (n = 25). Univariate analysis showed that Carotid Plaque-RADS, Carotid Plaque volume, and hypertension were associated with functional stenosis. After adjustment, Carotid Plaque-RADS remained an independent predictor (odds ratio: 2.35, p < 0.01) and demonstrated the strongest correlation (ρ = 0.51, p < 0.01). It also demonstrated good diagnostic performance (area under the curve [AUC]: 0.81; sensitivity: 85%; specificity: 68%) and favorable clinical utility on decision curve analysis. In an exploratory analysis, Carotid Plaque-RADS was also moderately correlated with CAD-RADS (ρ = 0.37, p < 0.01) and predicted CAD-RADS ≥ 3 with good discrimination (AUC: 0.72).

Conclusion: Carotid Plaque-RADS is an independent, noninvasive predictor of coronary functional stenosis in candidates for carotid revascularization.

Critical relevance statement: Carotid Plaque-RADS provides a noninvasive imaging-based tool that independently predicts coronary functional stenosis, thereby enhancing preoperative coronary risk stratification and supporting integrated cardiovascular management in candidates for carotid revascularization.

Key points: Carotid revascularization candidates face high coronary risk. Carotid Plaque-RADS independently predicts coronary functional stenosis. Carotid Plaque-RADS enhances preoperative coronary risk stratification.

Oncologic imaging plays a critical role in the diagnosis, staging, treatment planning, and follow-up of cancer patients. Recent advancements in computed tomography, particularly the development of photon-counting detector CT (PCCT), have introduced new opportunities for improving diagnostic accuracy and tissue characterization, while reducing contrast agent usage and radiation exposure. By offering ultra-high spatial resolution, enhanced contrast-to-noise ratio, and intrinsic spectral capabilities, PCCT addresses many limitations of conventional energy-integrating detector CT (EID-CT) and unlocks new possibilities for quantitative imaging. This review explores the emerging applications of PCCT across various tumor types-including thoracic, abdominal, and musculoskeletal malignancies-highlighting its potential to improve cancer imaging and patient care. CRITICAL RELEVANCE STATEMENT: Photon-counting detector CT (PCCT) offers several advantages in oncologic imaging, providing superior spatial resolution, spectral imaging capabilities, and reduced radiation dose, enhancing lesion characterization and precise treatment planning, making PCCT a valuable tool for personalized cancer care. KEY POINTS: CT has a crucial role in oncological imaging, supporting diagnosis, staging, treatment planning and follow-up. Compared to EID-CT, PCCT offers higher spatial and contrast resolution, reduces artifacts and image noise and provides spectral data enabling quantitative assessment. PCCT may improve cancer imaging by increasing diagnostic accuracy, with better detection of small lesions, enhanced soft tissue contrast, and enabling quantitative iodine uptake evaluation.

Objectives: To evaluate patient acceptance and feedback regarding supplemental imaging modalities: automated whole-breast ultrasound (ABUS), contrast-enhanced mammography (CEM), and abbreviated breast MRI (AB-MRI) within the BRAID (Breast Screening: Risk Adaptive Imaging for Density) trial.

Materials and methods: An adapted Testing Morbidities Index questionnaire was utilised to capture participant experiences and perceptions (January-April 2024) related to AB-MRI, ABUS and CEM. Likert-scale questions assessed discomfort, anxiety, and overall satisfaction for each imaging modality, while thematic analysis was applied to free-text patient feedback. Additionally, reasons for withdrawal were recorded for each modality.

Results: Among 159 women providing feedback, 57/159 (35.8%) underwent ABUS, 52/159 (32.7%) CEM, and 50/159 (31.5%) AB-MRI. Acceptability of ABUS, CEM and AB-MRI was rated similarly to mammography by 71/159 (64.8%) of these respondents, with 72/159 (45.3%) considering them superior. Mild-to-moderate discomfort due to breast compression was reported for ABUS and CEM, whereas AB-MRI resulted in the least discomfort. Pre-procedural anxiety was observed across all imaging modalities, particularly with contrast-enhanced techniques; however, experiences were generally well-tolerated. Effective communication and pre-test information reduced anxiety levels, with most participants willing to repeat the procedures. 151/984 (15.3%) withdrawals in BRAID were due to adverse patient experiences, with contrast-enhanced techniques accounting for most of these withdrawals (CEM: 69/151, 45.7%; AB-MRI: 66/151, 43.7%; ABUS: 12/151, 7.9%). The main reasons for withdrawal were unhappiness with the allocated imaging arm and discomfort or anxiety during the procedure.

Conclusion: Supplemental imaging modalities are generally well-accepted by patients with benefit throughout gained by clear communication and preparedness.

Critical relevance statement: Feedback from a subgroup of women participating in the BRAID trial shows that supplemental imaging alongside routine screening is well-accepted. Clear communication and empathetic care further improve acceptance, supporting a shift toward personalised breast cancer screening for women with dense breasts.

Key points: Understanding women's imaging experiences is essential for optimising breast screening practices. Acceptability of supplemental imaging was rated similar to or better than mammography by most participants. Clear, empathetic communication reduced anxiety and improved experience with contrast-enhanced imaging.

Prostate cancer is the most commonly diagnosed cancer among men in 112 countries, accounting for approximately 15% of all cancer cases. Whilst the 5-year survival rate for localised disease exceeds 90%, there is a significant drop to 50% if metastases are present. Following the VISION and TheraP trials, ¹⁷⁷Lu-PSMA-therapy was approved for treatment of metastatic castrate resistant prostate cancer by the FDA and EMA 2022. Patient selection for ¹⁷⁷Lu-PSMA-therapy is now relatively well defined, guided by PSMA-PET/CT criteria established in pivotal trials. Nevertheless, clinical consensus on appropriate criteria is still evolving, and additional imaging modalities such as ¹⁸F-FDG PET, post-therapy SPECT/CT, or emerging techniques such as whole-body diffusion-weighted MRI may serve as valuable adjuncts to identify PSMA-negative or treatment-resistant disease that may not be apparent on PSMA-PET/CT alone. This review examines the current evidence on imaging biomarkers and complementary diagnostic techniques used for patient selection, treatment monitoring, and response assessment in [¹⁷⁷Lu]Lu-PSMA-617 therapy for metastatic castrate resistant prostate cancer. Baseline imaging biomarkers on PSMA-PET/CT, such as mean standardised uptake value (SUV_mean), PSMA-avid total tumour volume, and inter-lesional PSMA heterogeneity, have shown promise in predicting treatment response and assessing outcomes. Additionally, statistical prognostic models have been developed to predict treatment efficacy, though further validation is required. Imaging plays a crucial role and should be considered alongside blood biomarkers, clinic-demographic history, and circulating tumour markers to improve patient selection for ¹⁷⁷Lu-PSMA-therapy. CRITICAL RELEVANCE STATEMENT: PSMA-PET/CT is the established imaging modality for patient selection for ¹⁷⁷Lu-PSMA-therapy, while ¹⁸F-FDG PET, post-therapy SPECT/CT, and emerging techniques such as whole-body diffusion-weighted MRI can be adjunctive for patient selection, response assessment and long-term monitoring. KEY POINTS: PSMA-PET/CT is the mainstay for patient selection for ¹⁷⁷Lu-PSMA-therapy. ¹⁸F-FDG PET, SPECT/CT or whole-body diffusion-weighted MRI could be used as adjuncts. Interim and longitudinal PSMA-PET/CT offer sensitive detection of progression, quantitative biomarkers for response assessment, and standardised frameworks. Advances in AI, radiomics, and standardisation frameworks may refine prognostication, enable personalised dosimetry, and integrate imaging biomarkers into clinical practice, though further validation is required.