Insights into Imaging最新文献

Objectives: High classification accuracy of Alzheimer's disease (AD) from structural MRI has been achieved using deep neural networks, yet the specific image features contributing to these decisions remain unclear. In this study, the contributions of T1-weighted (T1w) gray-white matter texture, volumetric information, and preprocessing-particularly skull-stripping-were systematically assessed.

Materials and methods: A dataset of 990 matched T1w MRIs from AD patients and cognitively normal controls from the ADNI database was used. Preprocessing was varied through skull-stripping and intensity binarization to isolate texture and shape contributions. A 3D convolutional neural network was trained on each configuration, and classification performance was compared using exact McNemar tests with discrete Bonferroni-Holm correction. Feature relevance was analyzed using Layer-wise Relevance Propagation, image similarity metrics, and spectral clustering of relevance maps.

Results: Despite substantial differences in image content, classification accuracy, sensitivity, and specificity remained stable across preprocessing conditions. Models trained on binarized images preserved performance, indicating minimal reliance on gray-white matter texture. Instead, volumetric features-particularly brain contours introduced through skull-stripping-were consistently used by the models.

Conclusion: This behavior reflects a shortcut learning phenomenon, where preprocessing artifacts act as potentially unintended cues. The resulting Clever Hans effect emphasizes the critical importance of interpretability tools to reveal hidden biases and to ensure robust and trustworthy deep learning in medical imaging.

Critical relevance statement: We investigated the mechanisms underlying deep learning-based disease classification using a widely utilized Alzheimer's disease dataset, and our findings reveal a reliance on features induced through skull-stripping, highlighting the need for careful preprocessing to ensure clinically relevant and interpretable models.

Key points: Shortcut learning is induced by skull-stripping applied to T1-weighted MRIs. Explainable deep learning and spectral clustering estimate the bias. Highlights the importance of understanding the dataset, image preprocessing and deep learning model, for interpretation and validation.

Objectives: To investigate normal and pathologic values of fatty infiltration (FI) and muscle volume through volumetric quantification of the main hip abductors of patients with unilateral greater trochanteric pain syndrome (GTPS) using 2-point-Dixon MRI.

Materials and methods: Patients prospectively underwent MRI of both hips: FI of the gluteus minimus (Gmin) and medius (Gmed) muscles were quantified by volumetric fat fractions (3D FF) using 2-point-Dixon MRI. Whole (WMV) and lean muscle volumes (LMV) were calculated for both muscles. 3D FF and volumes were compared between asymptomatic and GTPS hips, using the Wilcoxon signed-rank test. Gender-specific differences were assessed using the Mann-Whitney U test.

Results: Forty-one patients (mean age 65.0 ± 13.6 years, 27 females) were analyzed. 3D FF in asymptomatic hips was lower than in symptomatic hips (Gmin: 17.8% vs. 19.8%; Gmed: 12.7% vs. 15.9% (all p ≤ 0.02)). Gmin had a higher 3D FF than Gmed (p < 0.001). Females had higher FF (asymptomatic and symptomatic Gmin: 19.4%, 21.8%; asymptomatic and symptomatic Gmed: 13.2%, 16.3%) than males (asymptomatic and symptomatic Gmin: 14.7%, 16.1%; asymptomatic and symptomatic Gmed: 11.8%, 14.9%) for both sides and muscles. Average WMV in asymptomatic hips for Gmin and Gmed were 77.2 cm³, 270.1 cm³ in females, and lower in males (both p < 0.001) with 107.1 cm³, 408.1 cm³, respectively.

Conclusion: This study offers reference values for 3D FF and volumes of the Gmin and Gmed muscles in asymptomatic elderly hips, which are significantly lower than their GTPS counterparts, with succinctly higher fat fractions in females than males. Women showed significantly lower muscle volume for both muscles than men.

Critical relevance statement: Volumetric fat fractions of gluteal muscles show significant symptoms and gender related differences, indicating their potential as an imaging biomarker in the common GTPS patient.

Key points: In females, asymptomatic hips showed average volumetric fat fractions of 19% for Gmin and 13% for Gmed; with lower values in males, of 15% and 12%, respectively. Whole muscle volumes in asymptomatic hips for Gmin and Gmed were 77.2 cm³, 270.1 cm³ in females, and 107.1 cm³, 408.1 cm³ in males. Using volumetric fat fractions, abductor muscle fat content was significantly higher in symptomatic GTPS hips compared to asymptomatic hips.

Magnetic resonance imaging (MRI) is considered the gold standard for staging penile squamous cell carcinoma and assessing its extent. However, due to the rarity of this pathology, few medical centers have regular experience with penis carcinoma imaging. The purpose of this article is to provide a comprehensive update on the role of MRI in penile cancer by reviewing the MRI anatomy of a normal penis, outlining the recommended MRI techniques for penis assessment, and discussing the benefits and drawbacks of artificial erection. We will also highlight how MRI can serve the purpose of tumor staging and its therapeutic consequences. CRITICAL RELEVANCE STATEMENT: To provide a comprehensive and practical review of penile cancer based on imaging, including epidemiology, prognosis, treatment, penile MRI protocol, anatomy, and key points for accurate analysis. KEY POINTS: Penile carcinoma affects the glans and/or the foreskin in 98% of cases. MRI is the most accurate imaging modality for staging penile carcinoma and assessing its extent. T2-weighted using thin section is the best sequence to identify the tumor. Accurate treatment depends on the depth of local invasion and lymph node involvement.

Objectives: To evaluate the adjunctive diagnostic value of three-dimensional MR cholangiopancreatography (3D MRCP) for identifying biliary atresia (BA) in infants with cholestasis.

Materials and methods: This retrospective two-center study evaluated the adjunctive diagnostic performance of 3D MRCP beyond ultrasound (US) using receiver operating characteristic (ROC) analysis. The cohort from center 1 was divided into training (n = 770) and validation (n = 330) sets, with center 2 as the test set (n = 252). The optimal cut-off for the MR triangular cord thickness (MR-TCT) was derived from the area under the ROC curve (AUC) calculated from all cases. Extrahepatic bile ducts visualization on 3D MRCP was validated against surgical findings. Image quality metrics were assessed for their diagnostic value on BA detection.

Results: One thousand three hundred fifty-two eligible cholestatic infants undergoing 3D MRCP (February 2012 to June 2023) were enrolled, including 363 BA and 989 non-BA. ROC analysis identified 3.75 mm as the optimal cut-off MR-TCT for BA diagnosis (AUC = 0.828). The combination of MR-TCT, 3D MRCP, and US yielded superior diagnostic performance, achieving AUCs of 0.967 in the training set, 0.958 in the validation set, and 0.972 in the test set (all p < 0.001). Image quality scores (p < 0.001), signal-to-noise ratio (SNR) (p < 0.001), contrast ratio (p = 0.012), and contrast-to-noise ratio (CNR) (p < 0.001) of 3D MRCP significantly differed between correct and incorrect diagnosis groups.

Conclusions: 3D MRCP is a valuable diagnostic adjunct tool in diagnosing BA, particularly when combined with MR-TCT and US. Optimizing 3D MRCP image quality enhances diagnostic accuracy.

Critical relevance statement: 3D MRCP enhances BA diagnosis when combined with MR-TCT and US. Importantly, in cases with strong clinical suspicion but negative US findings, MRCP should be utilized as an adjunct diagnostic modality to reduce false-negative rates.

Key points: The diagnostic efficacy of 3D-MRCP in BA remains to be fully characterized. MR-TCT, 3D-MRCP, and US combined achieved optimal diagnostic accuracy for BA. For high-suspicion BA with negative US, adjunct 3D-MRCP reduces false-negative diagnoses.

Objectives: To evaluate the performance of semi-quantitative analysis versus quantitative pharmacokinetic analysis applied to ultrafast dynamic contrast-enhanced (UF-DCE) MRI in: (1) differentiating benign from malignant breast lesions, (2) molecular subtyping, and (3) predicting pathologic complete response (pCR) following neoadjuvant chemotherapy.

Materials and methods: This prospective noninferiority study enrolled 339 consecutive participants with suspected breast lesions between September 2022 and February 2024. All underwent breast DCE MRI with a temporal resolution of 4.5 s, totaling 100 phases. Using the initial 30 phases (early-phase UF-DCE), five semi-quantitative parameters were calculated: wash-in slope (WIS), time-to-peak, bolus arrival time, peak enhancement intensity (PEI), and initial area under the curve in 60 s. Furthermore, three quantitative parameters: volume transfer constant, rate constant (k_ep), and extravascular extracellular space volume, were derived from the 100-phase dataset (full time-course UF-DCE). Diagnostic performance was assessed using areas under the curve (AUC) and the DeLong test, with Pearson analysis evaluating the correlation between semi-quantitative and quantitative parameters.

Results: All semi-quantitative and quantitative parameters showed differences between benign and malignant breast lesions (p < 0.001). Semi-quantitative WIS demonstrated noninferior diagnostic performance to quantitative k_ep in differentiating benign from malignant lesions (AUC: 0.93 vs 0.92; ∆AUC = 0.02, p = 0.35). However, neither approach effectively distinguished molecular subtypes or predicted pCR (p > 0.05). Strong correlations were observed in PEI and K^trans (r = 0.75, p < 0.001).

Conclusion: Semi-quantitative analysis of early-phase UF-DCE exhibits noninferior performance to quantitative analysis of full time-course UF-DCE MRI for distinguishing benign from malignant breast lesions. Both analytical approaches showed limited utility in molecular subtyping and pCR prediction.

Critical relevance statement: Early-phase UF-DCE MRI provides a cost-effective alternative to full time-course UF-DCE MRI for differentiating benign and malignant breast lesions, demonstrating noninferior diagnostic performance with reduced scan time and no need for pharmacokinetic modeling.

Key points: Systematic comparison of early-phase UF-DCE and full time-course UF-DCE MRI for diagnosis, subtyping, and response prediction in a single breast cancer cohort remains limited. Early-phase UF-DCE MRI demonstrated noninferior diagnostic performance to full time-course UF-DCE MRI in differentiating benign and malignant breast lesions. Early-phase UF-DCE MRI is a time-efficient alternative to full time-course UF-DCE MRI for clinical implementation.

Objectives: To evaluate the technical parameters and clinical outcomes of contrast-enhanced mammography (CEM)-guided biopsy for diagnosing breast cancer in patients with suspicious lesions showing post-contrast enhancement on CEM but undetectable on standard digital mammography (DM) or ultrasound (US).

Materials and methods: A prospective study was conducted on 36 patients referred for CEM-guided biopsy based on suspicious (BI-RADS 4-5) enhancing-only lesions detected during previous CEM examinations and/or contrast-enhanced breast MRI (CE-MRI). Procedures were performed on a dedicated prone table using a vacuum-assisted biopsy device. Effectiveness parameters included success rate (lesion enhancement, diagnostic material collection, and correct clip positioning), procedure time, average glandular dose (AGD), compression force, and complication rate.

Results: From January to November 2024, 36 patients underwent CEM-guided biopsy, with a success rate of 97.2% (35/36). The median procedure time was 29 min. The AGD was 0.88 mGy (range 0.5-1.4 mGy, SD ± 0.22). The average compression force was 4.94 kg (range 2-7 kg, SD ± 1.1). Of the 35 lesions biopsied, 20 (57.1%) were masses and 15 (42.9%) non-mass enhancements, with a mean lesion size of 13.2 mm. Breast lesions were classified as BIRADS 4a (10/35), BIRADS 4b (5/35), BIRADS 4c (8/35), and BIRADS 5 (12/35). Histopathological findings showed 57.1% (20/35) of lesions were malignant. Lesion classification included 5.7% (B1), 34.3% (B2), 2.9% (B3), 31.4% (B5a), and 25.7% (B5b).

Conclusion: CEM-guided biopsy is an effective and accessible technique for targeting enhancing-only breast lesions, offering advantages over MRI-guided biopsy in terms of time, cost, and patient comfort.

Critical relevance statement: Contrast-enhanced mammography-guided biopsy is an effective and accessible technique for targeting enhancing-only breast lesions, offering advantages over MRI-guided biopsy in terms of time, cost, and patient comfort.

Key points: Contrast-enhanced mammography (CEM)-guided biopsy has recently gained traction as a reliable alternative to MRI for targeting enhancing-only lesions. This study explores the clinical implementation of CEM-guided biopsy in a prone position in our university hospital, assessing its efficacy, safety, and diagnostic accuracy. CEM-guided biopsy is a promising technique for the precise targeting of breast lesions, with advantages over MRI-guided biopsy.

Renal cell carcinoma is a prevalent malignancy affecting the urinary system and poses significant challenges in precision diagnosis and treatment. Although medical imaging technologies have been widely applied in renal cell carcinoma screening, traditional imaging diagnostics have limitations due to their high degree of subjectivity, relying primarily on the doctor's experiential judgment. The advent of radiomics presents a groundbreaking method for tackling this issue-by extracting high-throughput, deep-level information from conventional medical images to achieve a quantitative assessment of tumor characteristics. Furthermore, the fusion of radiomics and genomics has led to radiogenomics, which combines imaging features with molecular data, enabling the non-invasive evaluation of tumor biological behavior, molecular heterogeneity, and microenvironmental features, thereby providing a more detailed, accurate, and personalized assessment. In this review, we summarize the role radiomics and radiogenomics play in the diagnosis, prediction, and adjuvant treatment of renal cell carcinoma. Radiomics has demonstrated potential in classifying renal cell carcinoma subtypes, predicting patient prognosis, and forecasting disease progression. Radiogenomics further links imaging features to gene mutations and the tumor microenvironment, enabling non-invasive assessment of renal cell carcinoma biology and providing new approaches to diagnosis and treatment. CRITICAL RELEVANCE STATEMENT: By reviewing existing research, we summarize how radiomics and radiogenomics address key clinical challenges in the diagnosis and treatment of renal cell carcinoma, providing non-invasive solutions to overcome tumor heterogeneity and guide precision oncology. KEY POINTS: Renal cell carcinoma lacks reliable non-invasive biomarkers for precision diagnosis and characterization. Radiogenomics bridges imaging and molecular biology for precise predictions. Radiogenomics lacks full multi-omics integration despite data growth.

The increasing integration of artificial intelligence as medical devices (AIaMDs) within diagnostic imaging necessitates a robust understanding of associated regulatory frameworks among clinical practitioners. Despite the growing commercial availability and adoption of AIaMD, a significant awareness gap persists among radiologists regarding pertinent European Union regulations, including the Medical Device Regulation (MDR) and the novel EU AI Act, both of which lack explicit provisions tailored to AI components. This regulatory ambiguity underscores a critical need for clarified guidelines concerning "high-risk" AI classification and best practices for safe deployment within the radiological workflow. Legal responsibility for AIaMD Post-Market Surveillance (PMS) primarily rests with software providers, yet radiologists are expected to contribute to the ongoing monitoring of safety and performance. Recognizing the need to raise awareness and provide practical guidance, the European Society of Radiology (ESR) eHealth and Informatics Subcommittee, supported by the ESR AI Working Group, conducted a modified Delphi procedure involving 16 domain experts (of which 14 acted as panelists) to establish a set of shared recommendations. These aim to establish essential practices for AIaMD PMS and post-market clinical feedback (PMCF), as stipulated by the MDR and partially updated by the AI Act. This paper also provides an overview of relevant regulations to enhance awareness among all stakeholders, particularly deployers (e.g., radiologists) and providers (e.g., vendors). These recommendations represent a foundational step towards improving consistency in AIaMD deployment, providing a critical reference standard for physicians navigating the unique challenges posed by these novel technologies. CRITICAL RELEVANCE STATEMENT: Radiologists need to familiarize themselves with AIaMD EU regulations due to shared PMS responsibilities and current ambiguities. ESR recommendations aim to bridge this awareness gap, standardizing safe AI deployment and enhancing clinical feedback within medical imaging. KEY POINTS: Radiologists need a clear understanding of EU regulations for AIaMDs, as current laws lack imaging-specific guidance. There is a shared responsibility for AIaMD safety, with radiologists contributing to PMS and clinical feedback systems. The ESR provides crucial recommendations to standardize AI deployment and improve clinical feedback in imaging.

Breast screening reduces cancer-specific mortality but can also precipitate avoidable harms through over-detection of benign abnormalities and subsequent over-surveillance. Across mammography and digital breast tomosynthesis (DBT), ultrasound and magnetic resonance imaging (MRI), gains in sensitivity are often offset by reduced specificity, driving false-positive recalls, benign-biopsy burden and resource strain. Within breast imaging reporting and data system (BI-RADS)-guided decision-making, Category 3 and Category 4A trigger short-interval follow-up or biopsy despite low event rates, amplifying anxiety and cost. Artificial intelligence (AI) offers a practical route to mitigate these drawbacks. Prospective and real-world studies indicate that AI-assisted reading can maintain or improve cancer detection while lowering recall rates and workload. AI models also support finer risk stratification-particularly for BI-RADS 4 lesions-thereby reducing unnecessary interventions. This review synthesises evidence on the performance and limitations of mainstream screening technologies, delineates the multidimensional impact of over-detection, and evaluates the capacity of AI to rebalance sensitivity and specificity, optimise follow-up intervals and support risk-adapted workflows. A patient-centred, evidence-driven strategy that integrates validated AI with clearly defined decision thresholds and effective patient-provider communication can maximise benefit while minimising harm. CRITICAL RELEVANCE STATEMENT: This review critically evaluates the causes and consequences of over-detection and over-surveillance in breast cancer screening and highlights how AI can advance radiologic decision-making through improved lesion stratification and more efficient, personalised follow-up strategies. KEY POINTS: BI-RADS thresholds largely drive over-detection; refining downgrade rules for 3 and tightening biopsy in 4A may reduce unnecessary interventions without compromising cancer detection. Over-detection imposes burdens: unnecessary imaging and biopsies, psychosocial distress, economic costs, and environmental impact; its reduction enhances efficiency and patient safety. AI-assisted screening maintains or improves cancer detection while reducing recall rates and workload; it also enables risk-adapted management of BI-RADS 4A lesions, avoiding low-value procedures.

Objectives: To investigate the impact of deep learning reconstruction (DLR) on the image quality of diffusion-weighted imaging (DWI) for liver and its ability to differentiate benign from malignant focal liver lesions (FLLs).

Materials and methods: Consecutive patients with suspected liver disease who underwent liver MRI between January and May 2025 were included. All patients received conventional DWI (DWI_C) and an accelerated reconstructed DWI (DWI_DLR) in which acquisition time was prospectively halved by reducing signal averages. Image quality was compared qualitatively using Likert scores (e.g., lesion conspicuity, overall quality) and quantitatively by measuring signal-to-noise ratio of the liver (SNR_Liver) and lesion (SNR_Lesion), contrast-to-noise ratio (CNR), and edge rise distance (ERD). Apparent diffusion coefficient (ADC) values and diagnostic performance for differentiating benign from malignant FLLs were assessed.

Results: A total of 193 patients (128 males, 65 females; age range, 23-81 years) were included. For quantitative assessment, DWI_DLR demonstrated higher SNR_Liver, SNR_Lesion, CNR, and a shorter ERD (all p < 0.05). For qualitative assessment, DWI_DLR showed improved lesion conspicuity, liver edge sharpness, and overall image quality (all p < 0.01), with no significant difference in artifacts (p = 0.08). ADC values were lower with DWI_DLR for both benign and malignant FLLs (p < 0.001). In differentiating benign from malignant lesions, DWI_DLR achieved better diagnostic performance (AUC: 0.921 vs. 0.904, p < 0.05).

Conclusion: Deep learning-enhanced DWI enables a 50% reduction in acquisition time while simultaneously improving liver MRI image quality and diagnostic performance in differentiating benign from malignant FLLs.

Critical relevance statement: This study demonstrates that deep learning-based reconstruction enables faster, higher-quality liver MRI with improved diagnostic accuracy for focal liver lesions, supporting its integration into routine radiological practice.

Key points: Diffusion-weighted liver MRI commonly suffers from limited image quality and efficiency. Deep learning reconstruction substantially improves liver MRI quality while enabling significantly shorter acquisition times. Improved lesion differentiation enables more accurate clinical diagnosis of liver lesions.