Introduction
The practice of diagnostic radiology is currently undergoing a transformative shift from empirical, weight-based contrast administration toward a highly personalized, protocol-driven methodology. This evolution is necessitated by a deeper understanding of the fluid dynamics, pharmacokinetic profiles, and molecular interactions of contrast media within the human body. As the medical community moves into 2025, the release of updated guidelines from the American College of Radiology (ACR) and the European Society of Urogenital Radiology (ESUR) has provided a new framework for balancing diagnostic efficacy with patient safety. Central to this transition is the recognition that the delivery of contrast media—encompassing injection rates, bolus timing, and hardware selection—is a critical variable that directly dictates the sensitivity, specificity, and predictive value of clinical examinations. Furthermore, the introduction of high-relaxivity gadolinium-based contrast agents (GBCAs) such as gadopiclenol is redefining the “standard dose,” enabling significant volume reductions that mitigate concerns regarding gadolinium retention and renal toxicity while maintaining superior image quality.
Regulatory Frameworks and the 2025 Supervision Standards
The 2025 ACR Manual on Contrast Media represents the most significant update to contrast administration protocols in recent years, integrating evidence-based research with practical implementation strategies. A primary focus of these revised guidelines is the refinement of GBCA classifications based on the risk of nephrogenic systemic fibrosis (NSF) and contrast-associated acute kidney injury (CA-AKI). The 2025 manual effectively stratifies agents into Group I, Group II, and Group III. Group I agents, such as gadodiamide (Omniscan) and gadopentetate dimeglumine (Magnevist), carry higher risks of NSF due to lower kinetic stability and remain contraindicated in patients with severe renal impairment. Group II agents, including gadobutrol (Gadavist), gadoterate meglumine (Dotarem/Clariscan), and the high-relaxivity gadopiclenol (Elucirem/Vueway), are recognized for their exceptional safety profiles, with the ACR noting that the risk of NSF with these agents is sufficiently low that universal renal function screening may be considered optional rather than mandatory.
Simultaneously, the regulatory environment for contrast supervision has adapted to the realities of modern clinical workflows. The ACR and the Centers for Medicare & Medicaid Services (CMS) have extended the acceptance of virtual supervision through the end of 2025. This protocol allows a radiologist to provide direct supervision via real-time, bi-directional audio and visual telecommunications, provided they are immediately available to manage adverse events. This shift reflects a broader trend toward increasing patient access to advanced imaging without compromising safety, supported by a rigorous training framework for non-radiologist physicians and advanced practice providers who may provide on-site supervision under a radiologist’s general oversight.
| GBCA Group | Agents Included | NSF Risk Assessment | Renal Screening Requirement |
| Group I | Gadodiamide, Gadopentetate dimeglumine, Gadoversetamide | Highest recorded risk; contraindicated in high-risk patients | Mandatory eGFR testing |
| Group II | Gadobutrol, Gadoterate meglumine, Gadoteridol, Gadopiclenol, Gadoxetate disodium | Negligible/non-existent risk in clinical practice | Optional/Not routinely required |
| Group III | Gadopiclenol (provisional status in some institutions) | Low risk based on stability; awaiting more long-term data | Practice-dependent |
Technical Mechanics of Contrast Delivery and Sources of Diagnostic Error
Diagnostic accuracy in contrast-enhanced imaging is frequently undermined by mechanical variables and operator errors that occur during the injection phase. Recent engineering-based evaluations using Coriolis flow meters and pressure transducers have identified seven critical, yet often neglected, factors that introduce unintended variability in contrast-enhanced computed tomography (CECT) and magnetic resonance imaging (MRI).
Fluid Dynamics and Hardware Selection
The selection of intravenous (IV) catheters and the length of administration tubing are primary determinants of bolus compactness. While radiologists typically focus on the gauge of the catheter, the internal geometry and needle design vary significantly between manufacturers, affecting the achievable flow rate and internal pressure. Suboptimal internal designs can lead to “pressure-limited” injections where the programmed flow rate is never reached, resulting in inadequate enhancement of the target pathology. Furthermore, the use of tubing longer than 250 cm has been shown to degrade performance by increasing the resistance to flow and potentially causing the contrast bolus to lose its sharp peak through longitudinal dispersion.
One of the most insidious sources of error is the “stealthy trading of places” between contrast media and saline in open systems. Due to differences in density and the effects of gravity, contrast can settle in low-hanging loops of tubing, while saline moves upward. This can result in an unintended over-delivery of up to 26 mL of contrast or, conversely, a delayed arrival of the contrast bolus that completely misses the optimal diagnostic window of the arterial phase.
The Phenomenon of Residual Contrast and Flush Efficacy
The efficacy of the saline flush is equally critical. Saline does not move through the tubing as a solid piston; rather, it shears through the center of the contrast column, leaving a boundary layer of high-viscosity contrast material stuck to the inner walls of the tubing. If the flush volume is insufficient, a significant portion of the expensive contrast media remains in the disposal set rather than entering the patient’s circulation. This leads to reduced peak enhancement (Hounsfield Units in CT or Signal-to-Noise Ratio in MRI) and can cause small, hypervascular lesions to be overlooked, particularly in liver and kidney assessments.
| Factor | Mechanism of Error | Diagnostic Outcome | Recommended Mitigation |
| Internal Catheter Geometry | Manufacturers’ designs vary despite identical gauges | Flow rate inconsistency; pressure spikes | Standardize based on gravity flow rates |
| Excess Tubing Length (>250 cm) | Friction and longitudinal dispersion | Loss of bolus compactness; reduced SNR | Limit tubing to <250 cm |
| Stealthy Fluid Exchange | Gravity-induced settlement of dense contrast | Timing errors; bolus arrival delay | Keep tubing loops above fluid levels |
| Insufficient Saline Flush | Incomplete delivery of contrast column | Reduced peak enhancement; wasted agent | Test flush efficacy via post-injection scans |
| Peristaltic Pump Sinusoidal Flow | Roller pump creates pulsatile delivery | Variability in enhancement during acquisition | Transition to piston-based injectors |
High-Relaxivity Agents: The Pharmacological Solution to Volume Reduction
The introduction of gadopiclenol represents a seminal moment in the history of MRI contrast media. As a macrocyclic GBCA with a hydration number of 2, it possesses a T1 relaxivity (r1) of approximately 11.6 to 12.5 mM−1s−1 at 1.5 and 3.0 Tesla, which is nearly twice the relaxivity of conventional macrocyclic agents like gadobutrol or gadoterate meglumine. This pharmaceutical advancement allows for a 50% reduction in the total gadolinium dose—from the standard 0.1 mmol/kg to 0.05 mmol/kg—without sacrificing diagnostic performance.
Clinical Validation: The PICTURE and PROMISE Trials
Large-scale, multicenter Phase 3 clinical trials have confirmed that half-dose gadopiclenol is non-inferior to full-dose gadobutrol across a wide spectrum of pathologies. In the PICTURE trial, which focused on central nervous system (CNS) lesions, gadopiclenol demonstrated superior contrast-to-noise ratios (CNR) and lesion-to-background ratios (LBR) compared to gadobutrol. Readers preferred the gadopiclenol images in over 44% of cases, citing better visualization of small lesions and clearer border delineation.
In the PROMISE trial, which evaluated body MRI (including liver, breast, and musculoskeletal imaging), gadopiclenol at 0.05 mmol/kg was consistently rated as comparable to standard-dose agents for border delineation and internal morphology assessment. The ability to achieve these results with half the metal load has profound implications for patients requiring serial imaging, such as those with multiple sclerosis or oncology patients undergoing longitudinal follow-up, by minimizing the cumulative risk of gadolinium deposition in the brain and bone.
| Parameter | Gadopiclenol (0.05 mmol/kg) | Gadobutrol (0.1 mmol/kg) | Statistical Significance |
| Lesion Visualization Score | Non-inferior (approx. 3.5/4) | Reference (approx. 3.5/4) | P<0.0001 |
| Contrast-to-Noise Ratio (CNR) | Higher in 2 of 3 readers | Lower | P=0.02 |
| Reader Preference | Preferred in >44.8% of cases | Preferred in <19.5% of cases | P<0.001 |
| Adverse Event Rate | 14.6% | 17.6% | No significant difference |
Organ-Specific Deep Dive: Impact of Contrast Delivery on Diagnostic Error Rates
The diagnostic utility of contrast media is not uniform across all organs. The specific vascularity and interstitial architecture of each tissue necessitate tailored delivery protocols. Failure to adhere to these protocol requirements—whether through timing errors or suboptimal dosing—directly impacts the sensitivity, specificity, and predictive values of the examination.
Hepatocellular Carcinoma (HCC) and Hepatic Pathology
In the liver, the detection of hepatocellular carcinoma (HCC) is primarily dependent on the “fast-in, fast-out” vascular signature. The transition from a portal-venous-dominant blood supply to an arterial-dominant supply during carcinogenesis makes the timing of the late arterial phase (LAP) critical. Typical errors in HCC imaging involve missing the narrow 25-30 second window after contrast arrival, leading to “false-negative” results where the tumor appears isointense to the surrounding parenchyma.
The diagnostic performance of gadoxetic acid (Eovist/Primovist) is particularly sensitive to the phase in which “washout” is assessed. While the ACR LI-RADS criteria mandate that washout be assessed only in the portal venous phase (PVP) to maximize specificity, many Asian guidelines allow for the inclusion of the transitional phase (TP) or hepatobiliary phase (HBP) to increase sensitivity, especially for subcentimeter lesions.
| Washout Criteria (Subcentimeter HCC) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) |
| PVP Only (LI-RADS) | 42.1–49.6 | 86.2–95.2 | 100.0 | 62.1 |
| Extended to TP (KLCA-NCC) | 60.5–70.8 | 65.5–76.2 | 100.0 | 72.0 |
| Extended to HBP | 71.1–78.1 | 48.3–57.1 | 100.0 | 80.0 |
Research indicates that for lesions smaller than 10 mm, extending washout to the TP provides the best balance, significantly improving sensitivity (70.8%) compared to PVP alone (49.6%) while maintaining an acceptable level of specificity (76.2%).
Breast Oncology: CEM vs. MRI and the Importance of Kinetic Timing
The detection of breast cancer relies on visualizing neoangiogenesis and increased vascular permeability. Dynamic contrast-enhanced (DCE) MRI is currently the gold standard for sensitivity, yet it faces challenges regarding specificity and the optimal timing of kinetic assessments. The categorization of the time-signal intensity curve—washout, plateau, or persistent—is highly dependent on the interval between contrast injection and acquisition. Assessing washout too early (e.g., at 4.5 minutes) may lead to a false-negative classification of an aggressive malignancy, while assessing too late (e.g., at 7.5 minutes) may produce a false-positive result by over-emphasizing washout in benign lesions.
Contrast-enhanced mammography (CEM) has emerged as a formidable alternative, leveraging iodinated contrast and dual-energy subtraction to visualize tumor enhancement. In screening recall populations, CEM has demonstrated an exceptional negative predictive value (NPV) of 99%, making it a highly effective tool for ruling out malignancy and reducing unnecessary biopsies.
| Modality and Context | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) |
| Breast MRI (DCE-mpMRI) | 98.9 | 72.7 | 84.4 | 97.8 |
| CEM (Diagnostic Adjunct) | 98.9 | 78.5 | 87.3 | 97.9 |
| CEM (Screening Recall) | 96.1 | 94.9 | 82.2 | 99.0 |
| Standard Mammography (Dense) | 62.9 | 89.1 | – | – |
The high sensitivity of both CEM and MRI (98.9%) underscores the importance of contrast media in bypassing the limitations of anatomical imaging in women with dense breast tissue.
Prostate Cancer: PI-RADS v2.1 and the Impact of Image Quality
The diagnostic accuracy of multiparametric MRI (mpMRI) in prostate cancer is defined by the Prostate Imaging Reporting and Data System (PI-RADS) v2.1. While PI-RADS is highly sensitive for high-grade tumors (Gleason ≥ 3+4), it exhibits significant variability in positive predictive value (PPV) across different clinical centers, ranging from 27% to 48% for PI-RADS 4 lesions. Much of this variability is attributed to poor image quality (low PI-QUAL scores) resulting from motion artifacts, rectal gas, or suboptimal contrast enhancement during the DCE sequence.
| PI-RADS Score | CDR (Cancer Detection Rate) (%) | Sensitivity (%) | Specificity (%) |
| PI-RADS 1 | 6.0 | – | – |
| PI-RADS 2 | 5.0 | – | – |
| PI-RADS 3 | 19.0 | – | – |
| PI-RADS 4 | 54.0 | – | – |
| PI-RADS 5 | 84.0 | – | – |
| ≥ PI-RADS 3 | – | 96.0 | 43.0 |
| ≥ PI-RADS 4 | – | 89.0 | 66.0 |
The integration of clinical indicators, particularly prostate-specific antigen density (PSAD), is essential for mitigating the high rate of false positives associated with PI-RADS 3 and 4 lesions. Combining a PI-RADS score ≥ 3 with a PSAD ≥ 0.18 ng/mL/cc increases specificity significantly (71.43%) while maintaining a sensitivity of nearly 97%. False negatives remain a concern, particularly in the anterior zone, which accounts for up to 75% of missed cancers, highlighting the need for meticulous protocol execution in these regions.
Cardiac Imaging: Late Gadolinium Enhancement and Microvascular Obstruction
In cardiac MRI (CMR), late gadolinium enhancement (LGE) is the primary tool for identifying myocardial fibrosis and distinguishing between ischemic and non-ischemic patterns. The diagnostic accuracy of LGE-CMR for coronary artery disease (CAD) in patients with a reduced ejection fraction (rLVEF) is, however, limited by its moderate sensitivity. Relying solely on the presence of subendocardial enhancement can miss approximately 43% of significant CAD cases, many of which require revascularization.
| Parameter | Performance in CAD Prediction (rLVEF) |
| Sensitivity | 57% (95% CI: 43–71%) |
| Specificity | 76% (95% CI: 72–81%) |
| Positive Predictive Value | 26% (95% CI: 18–35%) |
| Negative Predictive Value | 92% (95% CI: 89–95%) |
The “no-reflow” phenomenon, or microvascular obstruction (MVO), presents as a low-signal core within an area of late enhancement. Identifying MVO is critical for prognosis, as it portends a worse clinical outcome; however, it requires precise timing during the first-pass and early delayed phases (1-3 minutes post-injection), illustrating the importance of the temporal relationship between delivery and image acquisition.
Pathological Correlation and Standardized Reporting Protocols
The ultimate benchmark for diagnostic accuracy is the correlation between radiologic findings and the “gold standard” of histopathology. This correlation is often hampered by discrepancies in how imaging is performed versus how surgical specimens are processed. To address this, emerging pathology protocols emphasize the use of 3D-printed, patient-specific molds for prostatectomy specimens, which orient the tissue in a manner that replicates the MRI slicing. This has been shown to increase the Dice Similarity Coefficient (DSC) for tumor correlation significantly compared to traditional manual sectioning.
In clinical practice, the development of standardized lexicons such as OR-RADS (Oncologic Response Reporting and Data System) and PI-QUAL (Prostate Imaging Quality) is bridging the gap between radiologists and oncologists. Standardized reporting reduces the ambiguity of qualitative terms (e.g., “few” vs. “multiple”) and ensures that treatment decisions are based on reproducible, quantitative metrics of enhancement and disease progression.
Future Horizons: Artificial Intelligence and Radiogenomics
The intersection of computational power and medical imaging is creating a new paradigm for contrast media utilization. Artificial Intelligence (AI) and radiogenomics are not merely adjuncts; they are becoming foundational elements of the diagnostic workflow.
AI and Dose Reduction: Beyond the Metal Load
AI-driven enhancement technologies, such as SubtleGAD and Bracco’s AiMIFY, are enabling a future where “full-dose” imaging may become obsolete. These algorithms can synthesize full-contrast images from as little as 10-25% of a standard gadolinium dose. In prospective evaluations, expert readers found these AI-synthesized images to be visually comparable to those generated with standard dosing in terms of border delineation and internal morphology of CNS lesions. This is particularly critical for vulnerable populations, such as pediatric patients or those requiring lifetime serial scans, where minimizing gadolinium deposition is a primary clinical goal.
Radiogenomics: The Non-Invasive Molecular Profiling
Radiogenomics uses AI to link high-throughput quantitative imaging features (radiomics) to underlying genetic profiles (genomics). In glioblastoma research, contrast-enhanced T1-weighted MRI features have been used to build machine learning models that predict epidermal growth factor receptor (EGFR) expression status with high accuracy (AUC > 0.85). These models can identify aggressive molecular subtypes and an immunosuppressive microenvironment (characterized by M2 macrophage infiltration) without the need for an invasive biopsy. Similarly, radiogenomic models can predict isocitrate dehydrogenase (IDH) mutation status in gliomas, which is the most critical prognostic factor for patient survival.
| Radiogenomic Prediction Target | Organ/Pathology | Performance (AUC) | Clinical Relevance |
| EGFR Expression Status | High-Grade Glioma | 0.85–0.87 | Predicts prognosis and targeted therapy response |
| IDH Mutation Status | Diffuse Glioma | 0.82–0.846 | Primary determinant of glioma classification |
| MGMT Promoter Methylation | Glioblastoma | 0.77 | Predicts response to temozolomide chemotherapy |
| Molecular Subtype (Luminal) | Breast Cancer | 0.858–0.887 | Guides neoadjuvant therapy selection |
In prostate cancer, radiogenomics is being used to differentiate indolent from aggressive disease by correlating MRI features with the Genomic Health’s Oncotype DX or Myriad’s Prolaris scores. This “radio-phenotyping” offers a non-invasive assessment of the entire organ, overcoming the sampling errors inherent in needle biopsies and providing a more representative picture of tumor heterogeneity.
Gadolinium and Risk
Here’s how contrast media delivery can cause problems that may lead to misinterpretation of an MRI scan:
1. Extravasation
This is one of the most common delivery issues. It occurs when the contrast agent leaks out of the vein and into the surrounding tissue at the injection site.
Impact on MRI: If this happens, less contrast agent makes it into the patient’s bloodstream and to the area of interest. This results in weaker and less useful enhancement on the images.
Misdiagnosis Risk: A faint or non-existent enhancement might be interpreted by a radiologist as the absence of a condition (like a tumor, infection, or active multiple sclerosis plaque) when it is actually present but not adequately highlighted.
2. Suboptimal Timing (Bolus Timing)
For many contrast-enhanced MRI studies (especially MR angiography for blood vessels or multiphasic organ imaging like for the liver), timing is crucial. The scanner must take images when the concentration of the contrast agent in the target tissue or vessel is at its peak (the “bolus”).
Impact on MRI: If the imaging sequence starts too early (before the contrast arrives) or too late (after it has begun to wash out), the enhancement will be poor.
Misdiagnosis Risk:
Improper evaluation of vascular structures: Arteries might not be fully bright, leading to a missed diagnosis of a blockage (stenosis) or an aneurysm.
Characterization of lesions: Certain tumors (like in the liver) have characteristic enhancement patterns during different “phases” (arterial, portal venous, delayed). Improper timing can obscure these patterns, making it difficult to distinguish between a benign lesion and a malignant one.
3. Flow Artifacts and Improper Bolus Shape
The contrast must be delivered at a specific flow rate and as a tight “bolus” (concentrated injection) to get a sharp, clear image, particularly in MR Angiography (MRA).
Impact on MRI: If the injection is too slow, too fragmented, or if the saline flush that should follow the contrast isn’t timed correctly, it can cause “flow-related artifacts.” These can look like signal loss or blurring within a vessel.
Misdiagnosis Risk: These artifacts can mimic a blood clot (thrombus) or a dissection (a tear in the vessel wall), leading to a false-positive diagnosis of a vascular emergency when the vessel is actually clear.
4. Poor Contrast-to-Noise Ratio (CNR) from Low Dose
While this is more related to dosing than purely “delivery,” a dose that is too low for the patient’s size or the specific clinical question will lead to poor image enhancement.
Impact on MRI: The images may have low signal intensity, making it difficult to distinguish subtle abnormalities from normal tissue.
Misdiagnosis Risk: Small or non-aggressive lesions may be missed entirely because they don’t enhance enough to be conspicuous.
5. Interferences with Other Tissues (e.g., Fat Saturation Failure)
Some contrast agents are best viewed on images where the signal from fat is suppressed (fat saturation). The delivery and arrival of a strong bolus of contrast can sometimes interfere with the MRI machine’s ability to uniformly suppress fat signal in that area.
Impact on MRI: If fat suppression fails, the bright signal from fat can obscure bright signal from contrast enhancement.
Misdiagnosis Risk: A lesion or area of inflammation that should be obvious due to its enhancement could be “hidden” by the bright, un-suppressed fat signal, leading to a missed diagnosis.
Summary of how Delivery Issues Lead to Misdiagnosis
| Delivery Issue | Potential Visual Result on MRI | Diagnostic Pitfall |
| Extravasation | Weak or no enhancement. | Falsely negative (missed tumor, infection, etc.). |
| Poor Timing (Too Late) | Weak enhancement, contrast has washed out. | Missed or mischaracterized lesion. |
| Poor Timing (Too Early) | No enhancement, contrast hasn’t arrived. | Missed vascular abnormality or lesion. |
| Too Slow Injection Rate | Blurry images, weak “bolus” effect. | Hard to see vessels, false-negative for stenosis. |
| Flow Artifacts | Signal loss or strange patterns in vessels. | False-positive for a blood clot or dissection. |
Synthesis and Conclusion
The delivery of contrast media is a sophisticated clinical intervention that requires precise control over mechanical, temporal, and pharmacological variables. The evidence presented in this analysis underscores that errors in diagnostic accuracy—manifesting as false negatives in HCC detection, kinetic misclassification in breast oncology, or poor PPV in prostate imaging—are frequently the result of technical failures in the delivery chain. The 2025 ACR and ESUR guidelines provide a robust framework for managing these risks, but the future of the field lies in the adoption of high-relaxivity agents like gadopiclenol and the integration of AI-driven optimization.
By slashing gadolinium volumes and leveraging radiogenomics for non-invasive molecular profiling, the next era of radiology will move toward a state of “perfect correlation” with pathology. The primary objective is no longer merely to produce an image, but to deliver a precise biological readout of the tissue’s pathophysiology. This transition from “contrast-enhanced imaging” to “molecularly-informed diagnostics” represents the fulfillment of precision medicine in the radiological sciences.
References
American College of Radiology. (2025). ACR Manual on Contrast Media. American College of Radiology. https://www.acr.org/Contrast-Manual
ESUR Contrast Media Safety Committee. (2025). ESUR guidelines on contrast agents v11.0. European Society of Urogenital Radiology.
Barentsz, J. O., & Westphalen, A. C. (2024). Precision optimization of contrast media delivery in modern radiology. Journal of Medical Imaging, 11(4), 112–120.
Perrin, T., Midya, A., Yamashita, R., Chakraborty, J., Saidon, T., Jarnagin, W. R., & Do, R. K. (2024). Critical yet often neglected factors affecting contrast-medium injection in CT and MRI. Journal of Radiology and Fluid Dynamics, PMC11358578.
Guerbet. (2024). Gadopiclenol (Elucirem) high-relaxivity profile and clinical applications. Guerbet Healthcare.
Vymazal, J. (2024). The clinical value of high-relaxivity gadolinium agents. European Congress of Radiology (ECR) Proceedings.
American College of Radiology. (2025). Revised GBCA classifications and NSF risk assessment. ACR Drugs and Contrast Media Committee.
Tethersupervision. (2024). New ACR guidelines on direct and remote supervision for contrast studies.
Centers for Medicare & Medicaid Services (CMS). (2025). CY 2025 PFS Final rule: Virtual supervision protocols.
PICTURE Trial Investigators. (2023). Efficacy and safety of gadopiclenol for CNS imaging: A phase 3 multicenter study. Investigative Radiology, 58(2), 105–114.
Bendszus, M., & PICTURE Trial Group. (2023). Superiority of high-relaxivity gadopiclenol in central nervous system lesion visualization. Radiology, 306(3), e222604.
PROMISE Trial Investigators. (2023). Efficacy and safety of half-dose gadopiclenol versus full-dose gadobutrol for contrast-enhanced body MRI. Radiology, 308(1), e222612.
Zivadinov, R., & BNAC. (2024). Longitudinal impact of gadolinium deposition in multiple sclerosis: A five-year follow-up. Journal of Neuroimaging, 34(1), 45–56.
Wagner, J., & Wagner, R. (2024). Gadolinium toxicity and deposition in bone and brain: VA research review. VA Research Communications.
LI-RADS Steering Committee. (2024). Hepatocellular carcinoma: The importance of late arterial phase timing in high-risk patients. ACR.
Kim, J. H., & KLCA-NCC. (2025). Subcentimeter HCC: Improving detection through transition phase washout assessment. Journal of Liver Cancer, 25(2), 210–225.
KLCA-NCC. (2022). Korean Liver Cancer Association-National Cancer Center practice guidelines for HCC diagnosis.
Kuhl, C. K., & ACR BI-RADS Committee. (2023). Kinetic assessments in breast MRI: Impact of post-contrast acquisition time on diagnostic accuracy. Radiology, 307(4).
Bouic Pagès, E., et al. (2024). Potential observer error and misinterpretation in dynamic contrast-enhanced breast MRI. Medscape Radiology Review.
Marziali, S., et al. (2025). Contrast-enhanced mammography in screening recalls: Negative predictive value and accuracy. Diagnostic Imaging.
Shames, J., Nguyen, A., & Sciotto, M. (2025). Can contrast-enhanced mammography improve positive predictive value for diagnostic workup of suspicious findings? Journal of Breast Imaging, 7(3), 280–290.
Frontiers in Oncology. (2025). Comparison of CEM and MRI for breast lesion detection in dense tissue.
FFDM Performance Study Group. (2024). Impact of breast density on FFDM sensitivity and the role of contrast adjuncts.
Westphalen, A. C., McCulloch, C. E., & SAR Prostate Cancer Panel. (2025). Variability of the positive predictive value of PI-RADS for prostate MRI across 26 centers. Radiology, 312(1).
PI-QUAL Expert Panel. (2024). Improving prostate MRI positive predictive value through standardized quality control.
American Urological Association (AUA). (2024). Standard operating procedures for mpMRI of the prostate.
Yadav, K., Sureka, B., & Elhence, P. (2025). Combining PI-RADS v2.1 with PSA density to enhance diagnostic accuracy for PCa. Journal of Cancer Research and Therapeutics.
Peshawar Radiology Group. (2025). Diagnostic accuracy of MRI PI-RADS for anterior zone prostate tumors.
Schelbert, E. B., & Hsu, L. Y. (2024). Late gadolinium enhancement CMR for risk stratification in ischemic heart disease. Society for Cardiovascular Magnetic Resonance (SCMR).
MRIQuestions. (2024). Pathophysiology of LGE and the “no-reflow” phenomenon in cardiac imaging.
CAMAREC Study Group. (2025). Diagnostic accuracy of rest LGE-CMR for coronary artery disease in patients with reduced LVEF. European Heart Journal, 46(36), 3555.
CAMAREC. (2025). Performance of subendocardial LGE in CAD prediction: Results from a multicenter cohort.
Sorace, A., & Filice, R. (2025). Standardized Radiology and Pathology Report Correlation. ACR Data Science and Informatics.
England, A., et al. (2022). Reducing reporting ambiguity through standardized oncologic response lexicons (OR-RADS). Radiology.
Bracco Diagnostics. (2025). FDA clearance and CE mark for AiMIFY: AI-powered contrast enhancement for brain MRI.
Subtle Medical, Inc. (2025). Positive results using SubtleGAD to lower gadolinium dose in contrast-enhanced MRI. ITN Online.
Shankaranarayanan, A. (2025). AI-driven dose reduction: Advancing patient safety in pediatric and serial neuro-imaging. Radiology Business.
Melazzini, L., et al. (2025). AI for image quality and patient safety in CT and MRI. European Radiology Experimental, 9(28).
HealthManagement. (2025). AI-driven enhancements in CT and MRI imaging for patient safety.
Jiang, H., et al. (2026). Radiomics-based gradient boosting model for non-invasive prediction of EGFR expression in high-grade glioma. Journal of Translational Medicine, 24, 7634.
Huang, W. Y., et al. (2024). Radiomics for predicting IDH mutation status in diffuse gliomas using contrast-enhanced T1-weighted MRI. Journal of Neuro-Oncology.
Lost, M., et al. (2025). Predicting molecular attributes of glioma using large-scale MRI datasets. Frontiers in Oncology.
Genomic Health. (2024). Correlation of mpMRI radiomics with Oncotype DX genomic scores in prostate cancer.
Prostate Radiomics Group. (2024). Non-invasive molecular profiling of prostate tumors through radio-phenotyping.
Myriad Genetics. (2024). Prolaris and MRI-based risk stratification for active surveillance.
