Current Progress in the CT- and MRI-Based Detection and Evaluation of Acute Pancreatitis Complications

Introduction

Acute pancreatitis (AP), one of the most prevalent gastrointestinal disorders, is characterized by premature activation of pancreatic zymogens triggered by etiological factors, such as gallstones, hypertriglyceridemia, and alcohol consumption [1]. This pathological process leads to localized pancreatic inflammation, which can progress to systemic organ dysfunction in severe cases. Globally, AP incidence varies widely from 3.4 to 73.4 cases per 100 000 individuals annually, with an observed annual growth rate of 2% to 5% [2, 3]. Clinical manifestations of AP are heterogeneous, ranging from self-limiting mild cases to life-threatening severe forms [4]. Approximately 25% of patients develop complications, including pseudocyst formation, pancreatic necrosis, peripancreatic fluid collections, or multiorgan failure, often requiring prolonged hospitalization, intensive care, and invasive interventions [5]. Cross-sectional imaging modalities, particularly computed tomography (CT) and magnetic resonance imaging (MRI), play a pivotal role in clinical practice by delineating pancreatic and peripancreatic structural abnormalities, thereby facilitating early detection of complications. Recent advances in radiomics further enhance diagnostic precision. For instance, a prospective study [6] by Tartari et al demonstrated that MRI-based radiomics (area under the curve [AUC]: 0.834) significantly outperformed CT alone (AUC: 0.610) in diagnosing acute pancreatitis, particularly in detecting early parenchymal edema on T2-weighted imaging, which is a precursor to complications such as acute necrotic collection (ANC) and acute peripancreatic fluid collection (APFC). This underscores the potential of quantitative imaging to augment conventional visual assessment. This review synthesizes recent advances in CT and MRI for identifying local and systemic sequelae of AP, aiming to enhance clinical decision-making and deepen the understanding of disease progression.

Localized Complications of AP

WALLED-OFF NECROSIS:

Walled-off necrosis (WON) can develop within 4 weeks following AP onset and is relatively common, being associated with various factors, including patient body mass index (BMI) [7]. Clinical choledochoscopic lavage is the primary approach to managing this condition, and patients face a risk of significant morbidity, infection susceptibility, and mortality [8,9]. The pathological presentation of WON is characterized by encapsulated fluid accumulation within and outside the pancreas. In this condition, the inflamed tissue wall remains largely intact and encircles, such that it is readily identifiable [10–12]. Contrast-enhanced CT demonstrates diagnostic value through heterogeneous non-enhancing areas (specificity: 80%–90%; sensitivity: 80–82.2%; Table 1) [13], while deep learning technology algorithms enable automated necrosis pattern analysis. Gupta et al [14] demonstrated that CT-based vision transformer models achieve superior performance in detecting >5-cm collections (sensitivity: 92.4–98.7%; specificity: 89.7–98.4%; AUC: 0.908–0.980). Diagnosing retroperitoneal fat necrosis, however, is hampered by the ability of CT imaging to reliably detect only necrosis, edema, and other pathological changes, while MRI can more reliably reveal retroperitoneal fluid and tissue necrosis [15]. Consequently, clinical evaluation should integrate patient-specific factors, such as BMI and fluid collection volume, to dynamically select artificial intelligence (AI)-enhanced CT or MRI modalities, enabling multimodal precision diagnosis.

ACUTE NECROTIC COLLECTION:

ANC is a condition that primarily results from necrotizing AP containing unequal amounts of fluid and necrotic material, and can involve the pancreatic and peripancreatic adipose tissues, without intact or recognizable wall formation. ANC can arise in many locations, including the abdominopelvic cavity, retroperitoneum, and pancreas [16]. Grassedonio et al [17] noted clear differences in the CT- and MRI-based detection of ANC, with CT images revealing inhomogeneous fluid density, whereas MRI results are characterized by an inhomogeneous signal intensity. During enhanced scanning, varying degrees of enhancement were evident for normal tissues, whereas the same was not true for necrotic tissues. Necrotic volume, the site of involvement, and numerous other factors influence whether or not the lesion fluid is absorbed.

Deep learning architectures, particularly convolutional neural networks (CNNs), enable automated quantification of ANC heterogeneity through voxel-wise analysis of density/texture features. In a multicenter validation study, Giri et al [18] demonstrated that CNN-based models achieved 91.3% accuracy (95% CI: 88.7–93.5%) in detecting ANC-specific imaging biomarkers, including CT hypodensity clusters (Δ >25 Hounsfield units) and MRI signal heterogeneity (coefficient of variation >30%). While MRI surpasses CT in resolving critical severity markers – such as pancreatic ductal discontinuities (sensitivity: 92% vs 68%), bubble signs (positive predictive value: 89% vs 62%), and micro-necrotic foci (<5 mm detection rate: 85% vs 45%) [19] – current deep learning implementations remain modality-specific. The integration of CT-derived radiomic features, such as gray-level co-occurrence matrix, with MRI functional sequences, such as diffusion-weighted imaging apparent diffusion coefficient maps, through multimodal deep learning frameworks could enable 4-dimensional spatiotemporal modeling of necrosis progression (κ=0.82 for inter-modality registration accuracy). Such an approach not only enhances ANC diagnostic accuracy (AUC improvement from 0.79 to 0.93 in pilot studies) but also predicts complication risks through necrosis resorption rate calculations (hazard ratio 2.1 for >30% volume persistence at 8 weeks). Despite these advances, 3 critical gaps persist: (1) lack of standardized protocols for cross-modal feature fusion, (2) insufficient external validation across diverse scanner vendors, and (3) absence of deep learning-driven intervention thresholds in clinical guidelines. Future research should prioritize developing vendor-agnostic neural networks trained on dual-energy CT/MR fingerprinting datasets, coupled with reinforcement learning algorithms to optimize intervention timing based on real-time necrosis volumetry.

ACUTE PERIPANCREATIC FLUID COLLECTION:

APFC typically develops within the first 4 weeks following the onset of AP and occurs exclusively in cases of interstitial edematous AP [18]. These fluid collections are located within the normal interfascial planes of the retroperitoneum and are not associated with peripancreatic necrosis. Pathologically, APFC manifests as a peripancreatic effusion, without the formation of pseudocysts, and most cases resolve spontaneously [20]. Studies have reported an incidence of APFC as high as 48.3%, highlighting its clinical relevance. Although APFC often resolves on its own, it can prolong hospitalization and pose significant risks to patients [21,22]. CT is a critical tool for evaluating APFC, as it provides high-resolution, multilayered images that delineate the lesion contour and differentiate liquefied or necrotic areas from surrounding normal tissues. Its precision makes CT invaluable for assessing disease progression in AP [21]. Similarly, Bhatia et al [23] noted that common sites of APFC include the small omental bursa and the left pararenal space. Over time, the fluid collections gradually resolve, typically achieving complete resolution within 5 weeks. While CT is widely used for diagnosing AP complications, it does have limitations. These include reduced sensitivity for differentiating ANC from APFC and limited ability to detect necrotic tissue fragments. MRI, on the other hand, offers superior soft-tissue resolution, enabling detailed visualization of solid necrotic components within fluid collections. MRI is particularly effective in distinguishing ANC from APFC. Emerging techniques such as diffusion-weighted imaging (DWI) hold promise for further advancing clinical practice by facilitating early detection of complications, such as infections, hemorrhages, or hematomas, as well as assessing fluid properties and monitoring treatment efficacy. These advancements are expected to significantly improve patient outcomes and management strategies in the future.

PANCREATIC PSEUDOCYST:

Pancreatic pseudocyst (PP) typically develops 2 or more weeks after the onset of AP and is influenced by factors such as hospitalization duration and alcohol intake [24]. Generally, the incidence of PP is estimated to be around 5% to 15%. However, this rate can be influenced by various factors, including the severity of pancreatitis, its underlying causes, and individual patient differences [25]. These cystic formations are most often located in the peripancreatic region and present with nonspecific symptoms, including nausea, vomiting, abdominal pain, and occasionally respiratory difficulties [26]. PP form when pancreatic enzyme secretion is dysregulated during an AP episode, resulting in the self-digestion of surrounding tissues and the extravasation of pancreatic fluid. As this fluid cannot be absorbed, it accumulates in the peripancreatic area and leads to the formation of rounded or elliptical cysts [27]. PPs lack solid content and contain minimal necrotic components. The cyst wall consists primarily of fibro-inflammatory tissue rather than epithelial tissue. In most cases, PPs are not connected to the pancreatic duct [28]. However, when the pancreatic duct becomes obstructed or stenosed, elevated ductal pressure can cause rupture or cystic dilation, leading to leakage of pancreatic fluid and the formation of PPs connected to the duct. In some instances, PPs can extend to the neck, causing difficulty swallowing [29]. On CT imaging, PPs typically appear as fluid-dense lesions with well-defined margins. Enhanced CT scans may reveal thick-walled, rounded peripancreatic fluid collections. While CT has demonstrated high diagnostic accuracy for identifying PP, with a specificity of 78.6% to 85.7% and sensitivity of 70% to 80% [13, 30], it has limitations, including the inability to reliably differentiate pseudocysts from pancreatic cystic tumors and the risk of contrast-induced nephropathy [31]. Habashi et al [32] reported that CT offers high diagnostic accuracy for PPs, surpassing that of abdominal ultrasound, allowing for the accurate evaluation of the prognosis of patients with AP. MRI, however, provides superior soft-tissue resolution and offers distinct advantages in characterizing PP [31]. On MRI, PPs are visualized as fluid accumulations with thin, symmetrical, and smooth walls. T2-weighted imaging excels in delineating the internal fluid components and structure of the cyst, while dynamic contrast-enhanced MRI can identify more complex features, such as cyst wall hypertrophy and vascular abnormalities [31]. In comparison to APFC, PPs tend to have a broader distribution and exhibit higher inflammation scores on CT. Severe cases of PP can require percutaneous drainage. During CT and MRI evaluations, careful attention should be paid to the distribution of peripancreatic fluid, its absorption, and its progression to assess the risk of PP formation. This approach provides a foundation for accurate clinical diagnosis and treatment. MRI offers more detailed information about the lesion’s characteristics and the effectiveness of treatment, while CT allows for rapid identification of anatomical features. In the future, AI may assist in the diagnosis of PP by analyzing imaging features, such as wall thickness and internal echoes. AI-driven tools may provide quantitative assessments that help differentiate pseudocysts from other fluid collections, ensuring that appropriate management strategies are implemented.

To provide a clearer comparison of the characteristics and onset times of these localized complications, Table 2 summarizes the key information regarding WON, ANC, APFC, and PP.

Systemic Complications with AP in CT and MRI Studies

ACUTE KIDNEY INJURY:

The kidney, as a retroperitoneal organ, is particularly vulnerable to inflammatory exudates caused by AP. When these exudates breach the perirenal fascial plane, they stimulate the perirenal space, impair renal function, and lead to alterations in biomarkers, such as calcitoninogen and triglyceride-glucose indices, potentially resulting in acute kidney injury (AKI) [33]. AKI complicates 14% to 16% of severe AP cases, with mortality risk increasing 3.2-fold in affected patients [34]. The retroperitoneal spread of pancreatic inflammation progresses through 3 anatomical pathways: (1) anterior pararenal space invasion, (2) perirenal fascial penetration, and (3) posterior pararenal compartment extension [34,35]. This process induces renal hypoperfusion (25–40% cortical blood flow reduction), interstitial hypertension (>20 mmHg), and hypoxic tubular damage [34, 35]. Contrast-enhanced CT serves as the primary diagnostic tool, quantifying perirenal fat stranding (CT attenuation >−30 HU) and fascial thickening (>3 mm). AI-enhanced protocols improve detection through radiomics analysis (Chen model [36] AUC 0.92) and perfusion mapping (medullary-to-cortical flow ratio <0.6 predictive of AKI) [37]. While bowel artifacts pose challenges, cycle-consistent generative adversarial networks reduce noise variance by 62% without compromising diagnostic features [38,39]. Current evidence confirms the renal safety of iso-osmolar contrast agents (eGFR decline <10 mL/min/1.73 m2) [40].

MRI’s multiparametric capabilities enable functional AKI assessment: blood oxygenation level-dependent (BOLD)-MRI detects medullary hypoxia (R2* >25 Hz), while diffusion-weighted imaging (DWI)/diffusion kurtosis imaging (DKI) identifies microstrutural damage (ADC <1.2×10−2 mm2/s, K >1.8) [41,42]. The integration of 3-dimensional CNNs with T2/DKI parameters achieves 91.7% severe AP-AKI prediction accuracy [43]. Emerging temporal AI models, such as long short-term memory networks, permit 55.8% early AKI detection through dynamic biomarker analysis [44]. Current challenges include further validation of the clinical applicability of DKI and the optimization of AI model generalizability for specific pathological subtypes, such as hyperlipidemic pancreatitis. Nevertheless, the synergistic advancement of AI and multimodal imaging is transforming AP-associated AKI diagnosis from passive recognition to proactive prediction, paving innovative pathways for personalized therapeutic interventions.

ACUTE RESPIRATORY DISTRESS SYNDROME:

The incidence of acute respiratory distress syndrome (ARDS) following AP typically ranges from 20% to 30%, particularly among patients with severe AP [45]. ARDS is one of the most dangerous complications of AP, with a mortality rate of approximately 50% [25]. Its pathophysiology is primarily associated with increased lung permeability and inflammation, driven by systemic inflammatory response syndrome. CT plays a crucial role in the evaluation of ARDS, as it can clearly reveal pulmonary abnormalities, such as fluid accumulation, solid lung lesions, and infiltrative changes [46,47]. Advances in imaging algorithms have enhanced the accuracy of lesion detection and accelerated the process of CT-based lung auto-segmentation for identifying solid lung lesions. MRI, while less commonly used for ARDS evaluation, offers unique advantages, due to its high sensitivity and superior ability to assess inflammation and edema in delicate lung tissues [48]. Techniques such as dynamic contrast-enhanced MRI and DWI allow for a more comprehensive evaluation of lung function impairment and edema in patients with ARDS. Although CT remains the primary imaging modality for ARDS assessment, MRI has potential as a complementary tool in complex cases that require more detailed tissue evaluation. Future research efforts may focus on integrating artificial intelligence, enhancing imaging technologies, and developing novel contrast agents to improve the diagnostic and prognostic capabilities of CT and MRI in the assessment of patients with ARDS.

CARDIAC INJURY:

Cardiac complications, including myocarditis, pericardial effusion, and heart failure, can arise as a result of systemic inflammatory response syndrome induced by AP. CT is commonly used to detect pericardial effusion or large-vessel abnormalities when evaluating cardiac complications [49]. However, its ability to assess myocardial function or detect subtle cardiac lesions is limited. Cardiac MRI can more comprehensively assess cardiac complications related to AP. It can detect myocardial edema, fibrosis, and necrosis through T2-weighted imaging and late gadolinium enhancement sequences [50]. Additionally, cardiac MRI provides detailed information on left and right ventricular function, allowing for a thorough assessment of overall cardiac performance [51]. This makes cardiac MRI particularly valuable in complex cases, as it delivers more precise diagnostic information than does CT. With continued technological advances, cardiac MRI is expected to play an increasingly important role in the diagnosis and management of cardiac complications associated with AP. Its ability to provide accurate, detailed assessments will enable physicians to improve diagnostic accuracy and tailor management strategies for patients with AP-related cardiac injuries.

LIVER INJURY:

The liver is often the first organ affected during the progression of AP, as it is highly susceptible to extrapancreatic injury. AP frequently involves multiple organ systems, and while the exact pathogenesis of liver injury in AP remains unclear, it is generally thought to be related to oxidative stress, inflammation, and other mechanisms [52]. On CT scans, liver injury is characterized by reduced parenchymal density and abnormal differentiation between fat, iodine, or calcium. These imaging features are particularly useful for distinguishing liver injury associated with AP from other liver disorders that also present with reduced density. Masuoka et al [53] reported that CT scans could reliably classify types of liver injury and detect associated biliary abnormalities in affected patients. However, Law et al [54] highlighted the limitations of CT, noting that liver injury is often subtle and prone to being overlooked or misdiagnosed based on imaging alone. A study by Fang et al [55], involving 144 patients with AP, demonstrated that the diagnostic accuracy of liver injury can be improved through detailed CT evaluation. This research underscored the clinical relevance of CT imaging in guiding treatment planning, particularly when combined with serum amylase levels for diagnostic analysis. Nevertheless, the role of CT in evaluating the severity of liver injury in AP remains controversial, and there is limited research exploring its application in this context. Additionally, CT has inherent limitations, particularly in assessing the severity of AP-associated liver injury. In contrast, the role of MRI in diagnosing liver injury in AP has been even less extensively studied. To date, there remains a lack of comprehensive research investigating the diagnostic potential of MRI in this setting. Future studies should therefore focus on exploring the utility of MRI for more accurately detecting and assessing hepatic damage in patients with AP.

SPLEEN INJURY:

Splenic injuries, most often presenting in the form of splenic vein thrombosis, splenic infarction, or splenic perforation, can also arise in patients with AP. These injuries typically develop due to elevated intra-abdominal pressure resulting from pancreatic inflammation and splenic vessel obstruction. Such complications are particularly prevalent in cases of necrotizing pancreatitis [56]. CT is the primary imaging modality for evaluating splenic injury, as it provides a clear depiction of the spleen’s anatomy and any lesions in the surrounding tissues. Enhanced CT is particularly effective as a means of assessing the patency of splenic vessels and identifying splenic infarctions or hemorrhages [57]. Although MRI has been used less frequently for the evaluation of spleen injury, it offers certain advantages. MRI can assess soft tissue structures and vascular patency in higher resolution. Dynamic contrast-enhanced MRI, in particular, is valuable for evaluating splenic perfusion and detecting localized infarcted or necrotic areas [58]. While CT remains more widely used in this context, MRI is expected to gain prominence in the future as advancements in technology and further research enhance its utility. With its superior soft tissue resolution, MRI can provide detailed insights into splenic tissue abnormalities and related conditions, such as splenomegaly, necrosis, or peri-splenic fluid accumulation, offering complementary diagnostic information to that provided by CT.

PANCREATIC ENCEPHALOPATHY:

Pancreatic encephalopathy (PE) is a relatively rare complication of AP, with a low incidence compared to other AP-associated complications. However, when PE occurs, it is associated with a significantly higher mortality rate of up to 43%. PE is characterized by neurological dysfunction, including cognitive impairment, seizures, and altered consciousness [59]. CT offers value as a means of excluding structural brain lesions, such as hemorrhages or infarcts. However, its sensitivity when detecting functional brain injuries is limited. In contrast, MRI scans, particularly DWI sequences, are highly effective in identifying early cerebral white matter edema and microscopic lesions associated with PE [60]. Additionally, BOLD-MRI can assess cerebral blood flow and oxygenation, providing valuable functional information that may aid early intervention efforts [61]. Despite the potential of these imaging techniques, there is currently limited clinical research on PE, with most studies focusing on serum biomarkers such as interleukin-6 and tumor necrosis factor-α. Future investigations should further explore the role of advanced MRI sequences, including BOLD and DWI, to enhance the early diagnosis and management of PE.

In addition to causing localized complications, AP can lead to significant systemic complications that can impact multiple organ systems. To provide a comprehensive overview of these systemic effects, Table 3 summarizes various complications, their onset times, and key characteristics.

FUTURE DIRECTIONS:

As imaging technologies continue to evolve, the synergistic integration of CT and MRI is poised to significantly enhance early diagnosis and clinical management of AP complications. This approach is supported by clinical evidence; for instance, Tartari et al reported [6] that a combined CT-MRI radiomics model achieved better diagnostic performance (AUC 0.911) for AP than did unimodal analyses, demonstrating the added value of integrated multimodal data. In addition, AI-assisted diagnosis of AP complications is an important future direction, which will greatly improve accuracy, such as for the automatic identification of necrotic areas [14,15] and automatic quantification of voxels for density/texture features [18,19]. However, the current model does not completely cover AP complications, and the efficacy of some AI-assisted diagnosis is not obvious; therefore, in-depth research in this area is needed.

Conclusions

CT remains the preferred modality for initial assessment of AP complications, leveraging its high spatial resolution and rapid acquisition to effectively detect local sequelae, including peripancreatic fluid collections, necrosis, and pseudocysts. However, CT demonstrates limited capability in discriminating necrosis subtypes or evaluating functional deficits. MRI offers critical complementary advantages through superior soft-tissue contrast and functional assessment, enabling detailed visualization of pancreatic ductal anatomy while precisely characterizing soft-tissue abnormalities, inflammatory activity, and perfusion dynamics. Consequently, selecting the optimal imaging modality based on specific complication patterns is essential for guiding clinical management and improving patient outcomes.