Computed tomography quantitative imaging features for pancreatic ductal adenocarcinoma after neoadjuvant therapy: a narrative review

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal solid malignancies. The essential problems of early systemic spread and aggressive biology have not been solved, despite improvements in survival (1). In recent years, neoadjuvant therapy (NAT) has become the standard treatment strategy for borderline resectable and locally advanced PDAC, increasingly used in selected resectable disease. The goal is not to pursue the reduction of tumor on the image, but to treat occult micrometastases early and to test whether tumor biology needs morbidity-intensive surgery (2-5). During therapy-driven change, as the use of NAT expands, imaging is expected to function less as an anatomic inventory and more as a decision tool for timing, resectability, and prognosis. Given the key role of computed tomography (CT) in staging and surgical decision-making worldwide, this review is intentionally CT-based, with magnetic resonance imaging (MRI) and positron emission tomography (PET) discussed only as complementary parts.

The post-NAT question is clear: is there surgically meaningful residual disease, especially at the tumor-vessel interface and what is the likelihood of a margin-negative (R0) resection? CT assessment is intrinsically challenging because fibrosis, edema, and inflammation can mimic residual tumor, while viable nests may persist without substantial size reduction (6-8). In current clinical practice, radiologists rely on enhancement patterns, vascular contact, and peripancreatic soft tissue changes, yet these surrogates show nontrivial inter-reader variability and do not consistently yield reproducible prognostic stratification across centers (6). Appearances can be described, but it is difficult to accurately quantify the clinical implications. This limitation has also prompted the development of pancreas-specific response frameworks, such as pancreatic Response Evaluation Criteria in Solid Tumors (pRECIST), suggesting the need for more standardized and disease-tailored approaches to response assessment in pancreatic cancer (9).

This inconsistency is mechanistic rather than accidental: it reflects the PDAC tumor microenvironment (TME). PDAC is characterized by dense desmoplastic stroma, abnormal microvasculature, and hypoxia. NAT can remodel this ecosystem in clinically meaningful ways that influence outcomes without producing a clean morphologic response (10-12). Even though these processes influence CT phenotypes and prognosis, conventional visual criteria are poorly suited to tracking stromal reorganization, collagen deposition, and shifts in immune cell infiltration. CT quantitative imaging features attempt to solve these biology-driven ambiguities. Radiomics and delta radiomics quantify heterogeneity and temporal remodeling; peritumoral/perivascular analysis focuses on the clinically decisive “invasion versus scar” interface; and energy-resolved CT provides material-sensitive readouts that may improve the accuracy of interpreting post-treatment results (13-17).

Accordingly, a CT-based, microenvironment-aware decision support framework is proposed for PDAC after NAT to make post-therapy imaging more reproducible and more actionable for multidisciplinary team (MDT) decisions (6,14,16,17). Specifically, quantitative CT readouts are organized around four critical MDT decisions: (I) resectability and the tumor-vessel interface; (II) whether to continue, escalate, or modify NAT; (III) timing the surgical window; and (IV) early recurrence risk stratification (Figure 1). To operationalize this framework, pathologic response and therapy-driven TME remodeling are linked to CT phenotypes, synthesizing evidence across conventional multiphasic quantification, perfusion-oriented metrics, and energy-resolved techniques. In parallel, key methodological requirements for center-spanning deployment are outlined, including protocol discipline, segmentation governance, feature stability testing, external validation, and calibration (16,18). We present this article in accordance with the Narrative Review reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0239/rc).

Figure 1 Decision-oriented CT reporting framework for post-NAT PDAC. CT, computed tomography; MDT, multidisciplinary care; NAT, neoadjuvant therapy; PDAC, pancreatic ductal adenocarcinoma; R0, margin-negative resection.

Methods

A narrative review was conducted to synthesize CT-based quantitative imaging features for PDAC after NAT, with emphasis on resectability assessment, NAT stewardship, surgical-window timing, and early recurrence risk stratification. In order to update the final draft before submission, an updated primary search was performed on 7 April 2026 in PubMed and Web of Science Core Collection to identify English-language articles published from 2013 to 2026. The search focused on PDAC, NAT, CT, and CT-derived quantitative approaches, including radiomics, perfusion CT, dual-energy/spectral CT, and photon-counting CT (PCCT), together with clinically relevant endpoints such as response, resectability, margin status, survival, recurrence, and prognosis. Targeted supplementary retrieval and manual identification were used to identify and incorporate recent topic-relevant references supporting post-NAT CT assessment, biomarker integration, standardized anatomic reporting, quantitative imaging methodology, complementary MRI and PET evidence, and methodological framework literature. Titles and abstracts were screened for relevance, followed by full-text assessment of potentially eligible records. A summary of the search strategy is provided in Table 1, and the detailed search strategy is provided in Table S1.

Pathologic staging, TME, and imaging correlates

After NAT, PDAC often shows a pattern of “response by replacement”. It refers to a decrease in viable tumor glands, but the treated mass is often dominated by therapy-induced fibrosis, necrosis, edema and inflammation (6-8). As a result, the apparent mass can persist despite meaningful biological response. This pathology-imaging inconsistency is a key reason why size-based criteria may underestimate treatment effect. The CT image of post-NAT is interpreted as a readout of residual tumor plus microenvironmental remodeling, rather than tumor cell density alone.

Pathologic tumor regression grading (TRG) systems, including College of American Pathologists Tumor Regression Grading System (CAP) and Evans Tumor Regression Grading System (Evans), quantify the residual viable tumor relative to fibrosis (19-21). However, interobserver variability and cross-system nonequivalence limit their use as a universal “gold standard” endpoint across cohorts. With the development of imaging biomarkers, this has a practical significance: models should be assessed not only against TRG, but also against clinically decisive endpoints and externally validated across centers to ensure robustness (22).

Mechanistically, the dominant substrate is the PDAC TME. A collagen-rich desmoplastic stroma—driven by cancer-associated fibroblasts (CAF)/stellate cell programs and extracellular matrix (ECM) remodeling—creates a stiff, hypoxic milieu so that it impairs drug delivery and changes enhancement kinetics. Importantly, stromal biology is heterogeneous, and different CAF/ECM programs may exert distinct effects on invasion and treatment sensitivity (10,23-25). At the same time, PDAC is often immune excluded because of enrichment of immunosuppressive myeloid/treg populations, although NAT can partially reprogram immune contexture (11,12,26,27). From an imaging perspective, these immune inflammatory changes are more likely to manifest on CT at the peritumoral compartment and tissue interfaces than as tumor alone (15,28).

Accordingly, post-NAT CT image can be framed along three related axes throughout this review: immune inflammatory activity, stromal remodeling, and hypoxia-vascular alteration (13-17). Delta radiomics captures temporal remodeling (16); peritumoral/perivascular gradients quantify interface biology (15); perfusion metrics operationalize vascular function (29,30); and iodine/material-sensitive measures on spectral and PCCT improve biological specificity and reproducibility. In general, these microenvironment linked readouts are positioned to support four critical MDT decisions: resectability at the tumor-vessel interface, whether to continue/modify NAT, timing the surgical window, and early recurrence risk stratification (6,30-33). This framework is summarized as a pragmatic CT-centric measurement ladder in Table 2.

Quantitative imaging workflow and methodological considerations

Robust quantitative imaging after NAT requires discipline across the entire analysis pipeline, from acquisition to validation (Figure 2). The first step is image acquisition and reconstruction. For CT, this typically involves a dedicated pancreatic protocol with thin-slice multiphasic imaging and standardized contrast timing, ideally including late arterial or pancreatic phase and portal venous phase. Small changes in slice thickness, reconstruction kernel or noise reduction algorithms can substantially alter the distribution of radiomic features (16). Harmonization techniques (e.g., feature standardization or ComBat-based adjustment) may reduce interprotocol variability. However, if feasible, prospective protocol optimization and adherence to Image Biomarker Standardization Initiatives (IBSI) are preferable (16). In longitudinal post-NAT assessment, image registration may further improve temporal comparability across serial CT examinations by stabilizing lesion matching and objective response measurement, particularly when tumor margins become ill-defined during therapy (41).

Figure 2 Schematic workflow for advanced CT quantitative analysis in post-NAT PDAC. The schematic summarizes the sequential process from pancreatic CT acquisition and baseline-restaging registration, through segmentation and compartment-aware feature extraction, to model development and validation. Key safeguards highlighted in the workflow include protocol stability, compartment-aware analysis, feature reproducibility, and clinical robustness. The model validation step further emphasizes calibration, clinical utility assessment, and overfitting control. CT, computed tomography; NAT, neoadjuvant therapy; PDAC, pancreatic ductal adenocarcinoma; ROI, region of interest.

Segmentation is the second critical step. Many PDAC radiomics studies perform manual or semi-automatic delineation of the primary tumor on a single pancreatic phase CT (15-17). More sophisticated protocols explicitly contour perivascular regions, peripancreatic fat and adjacent organs, sometimes using distance-based expansions around the gross tumor volume to capture the peritumoral compartment. Habitat analysis segments tumors to subregions according to intensity or texture clustering, generating “core”, “rim” or other biologically motivated habitats (15,69).

Once regions of interest (ROI) have been defined, feature extraction can proceed. Hand-crafted radiomics typically produces descriptors of shape, first-order statistics, and texture features, calculated on single time points or derived as delta radiomic features that explicitly encode changes between baseline and post-NAT scans (16). Deep learning models learn hierarchical features directly from image patches or volumes, with or without explicit segmentation, and hybrid models that fuse hand-crafted and deep features may improve prognostic performance in PDAC by capturing complementary aspects of tumor heterogeneity and stromal architecture (17). Although this review is CT-centric, MRI or PET may provide supplementary information when post-NAT CT is biologically equivocal (8,70,71).

The final step is model development and validation. At present, small sample sizes, particularly in single-center post-NAT series, make overfitting a constant risk. Increasingly, studies adhere to methodological frameworks such as the Radiomics Quality Score (RQS), Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) and the Prediction Model Risk of Bias Assessment Tool (PROBAST), which can promote prespecified analysis plans, reporting of calibration and assessment of clinical utility (18,59). Without these safeguards, apparently impressive area under the curve (AUC) in retrospective studies may not translate into robust, clinically usable tools.

Conventional quantitative CT

Pancreatic multiphasic CT is the primary imaging method for baseline staging, restaging, and operative decision-making in PDAC treated with NAT. However, the post-treatment pancreas is a biologically “noisy” environment: even though the size of the soft-tissue mass is stable or larger, it may still coexist with meaningful tumor regression because therapy-induced fibrosis, edema, and inflammatory change often persist even when viable tumor has significantly decreased. This pathology-imaging mismatch underlies the modest correlation between CT assessment in PDAC after NAT and pathologic response and survival. In practice work, CT is frequently reflecting the mixed state of residual tumor and reconstructed stroma, not just the density of tumor cells (6,31).

This pathology-imaging mismatch does not weaken the role of standardized anatomic reporting. Instead, it defines standardized reporting as the anatomical scaffold for subsequent quantitative CT refinement. The pancreatic cancer radiology reporting template recommended by the National Comprehensive Cancer Network (NCCN) and the Society of Abdominal Radiology/American Pancreatic Association consensus reporting template provide a common language for lesion location, pancreatic ductal dilatation, biliary obstruction, arterial/venous contact, venous contour irregularity or narrowing, loss of fat planes, and extrapancreatic disease (34,35). These elements define the anatomic map used by surgeons and MDT decisions to judge resectability and plan exploration (34). However, in PDAC after NAT, this map should not be treated as a sufficient response biomarker (36). Arterial/venous contact, vessel encasement, loss of fat planes, and perivascular soft tissue are essential to report, but therapy-induced fibrosis, edema, inflammation, and stromal scarring can reduce the predictive value of morphology-based assessment for vascular invasion (6,37). Therefore, standardized anatomic reporting and quantitative CT should be integrated hierarchically. The reporting template defines where disease is and which vessels or compartments are at risk, whereas longitudinal enhancement, tumor-vessel interface change, delta radiomics, iodine/perfusion metrics, and habitat-level features refine what the residual soft tissue may represent and how it may influence MDT decisions after NAT. Similarly, regional lymph-node assessment may be handled in the same hierarchical manner. Quantitative CT features of suspicious nodes, including size, morphology, enhancement, and texture- or radiomics-derived descriptors, may provide complementary information for preoperative MDT assessment, although direct validation in post-NAT PDAC cohorts is limited (38).

A pragmatic approach is to operationalize routine CT descriptors as measurable phenotypes, especially at the tumor-vessel interface where MDT decisions are made. Morphologic features such as tumor volume, contour sharpness, loss of fat planes, and patterns of arterial/venous contact can be quantified systematically rather than described impressionistically. Recent studies have provided concrete examples of how routine CT morphology can be converted from visual impressions into measurable CT phenotypes. For example, pancreatic surface lobularity combined with attenuation thresholds has been used to quantify parenchymal morphology, such as histologic fatty infiltration, and the computer-assisted edge analysis has been used to quantify pancreatic margin irregularity and distinguish PDAC from normal pancreatic parenchyma (39,40). Although these studies were not performed in post-NAT PDAC cohorts, they may support the feasibility of converting contour sharpness, surface lobularity, attenuation-based descriptors, and edge features into measurable CT phenotypes. As a result, these methods may serve as a transparent baseline layer of quantification in the CT-centric measurement ladder, before more complex delta radiomics, habitat analysis, or artificial intelligence (AI) models are applied. In the post-NAT setting, their extension may help characterize treatment-related changes in tumor contour, attenuation trajectories, boundary definition, and tissue interfaces. Although the 180-degree vessel contact threshold is useful for baseline anatomic staging and structured communication, its independent predictive value decreases after NAT, because perivascular soft tissue may reflect residual tumor, stromal scarring, or interface inflammation. Post-NAT vascular contact should not be interpreted as a binary marker of residual vascular invasion or R0 probability. Instead, arterial/venous contact category may be integrated with longitudinal interface change, enhancement or attenuation trajectories, perivascular soft-tissue evolution, and clinical variables, including carbohydrate antigen 19-9 (CA 19-9) and treatment tolerance (6,37). Together, if microenvironmental factors are ignored and only morphology is interpreted, judgments regarding tumor resectability may vary across readers and centers (15,37). Trajectory-based quantification can help define a surgical window. Stability or improvement in interface focused metrics despite limited size change may support proceeding to exploration, whereas deterioration can justify extending or adapting NAT before committing to high-morbidity surgery.

Attenuation and enhancement may likewise be treated as quantitative surrogates, not only visual impressions. PDAC is classically hypoenhancing in the pancreatic phase, whereas after NAT, enhancement commonly becomes heterogeneous, reflecting spatial mixtures of viable tumor nests, collagen-rich fibrosis, and altered microvascular function. A small set of interpretable metrics, such as tumor-to-parenchyma enhancement ratios across phases, histogram based heterogeneity indices, and intensity distribution descriptors, can serve as a transparent baseline layer of quantification. In addition, extending these measurements to longitudinal deltas is biologically aligned with the NAT setting, because it encodes treatment-induced remodeling rather than baseline phenotype alone (14). Longitudinal changes in enhancement- and heterogeneity-based metrics may help guide NAT stewardship particularly when conventional morphology is equivocal. However, these imaging trajectories should not be interpreted in isolation. In clinical MDT practice, they need to be integrated with serum CA 19-9, treatment tolerance, biliary status, and overall performance status, because the discordance between radiologic morphology and biochemical response may indicate residual systemic risk or treatment-related imaging ambiguity (6,42-44).

Beyond the tumor mass, interface focused quantification is often most decisive for MDT decisions. Radiomics applied to perivascular regions has been used to capture subtle changes in attenuation and organization of perivascular fat and small venous collaterals. Early studies suggest these features may improve prediction of true vascular invasion and R0 probability beyond diameter-based criteria (15,37,72). Similarly, peripancreatic fat stranding—often dismissed as nonspecific—may also encode quantifiable information and fibrotic responses that relate to local control and recurrence patterns (15,37). Integrated with clinical variables, these interface and peritumoral signatures are well-suited to stratify early recurrence risk. Thereby, these image features directly link conventional CT quantification to MDT planning for surveillance intensity and adjuvant strategy (15).

Functional CT

Functional CT perfusion reframes the question after NAT from morphology toward microvascular delivery—how blood flow and iodinated contrast reach the treated tumor bed. Dynamic acquisitions yield time-attenuation curves, enabling either kinetic estimates (e.g., blood flow, blood volume, mean transit time, permeability-related terms) or pragmatic semiquantitative surrogates (peak enhancement, time-to-peak, upslope) when full modeling is impractical (45).

From a biological standpoint, these readouts relate most directly to the microenvironmental axis of stroma-hypoxia-vasculature: dense desmoplasia and elevated interstitial pressure reduce effective perfusion, whereas therapy-induced stromal reorganization or vascular reprogramming can alter curve shape even when lesion size remains stable (10,29,45). In PDAC, CT perfusion typically shows lower perfusion than nontumoral pancreatic parenchyma, supporting the concept that perfusion metrics can serve as alternative surrogates of a constrained microvascular environment (29).

From a clinical standpoint, perfusion could be most defensible as a selective method at high-impact MDT inflection points rather than a universal restaging tool. When routine CT is equivocal, particularly at the tumor-vessel interface, trajectory-based changes in perfusion metrics can complement conventional phenotyping (46). In practice, the additional value is greatest when the readouts are explicitly tied to MDT decisions (6,30,31). Evidence in NAT-treated cohorts is still modest, but initial experience suggests perfusion CT may help predict histopathologic response in selected settings (30).

Adoption is constrained by dose, motion sensitivity and model heterogeneity, which limits cross-center interchangeability. When full perfusion CT is impractical, energy-resolved CT may provide more scalable surrogates, such as iodine/material-sensitive quantification and improved contrast-to-noise with PCCT, to probe related vascular biology within routine multiphasic workflows (32).

Dual-energy and spectral CT

Energy-resolved CT augments the standard pancreatic multiphasic protocol with quantitative channels from anatomy toward measurement. Spectral reconstructions—virtual monoenergetic images (VMIs), iodine maps, effective atomic number, and virtual noncontrast—allow routine CT to be interpreted as a set of measurable phenotypes rather than a single grayscale snapshot. In post-NAT PDAC, where fibrosis, edema, and inflammation narrow the dynamic range of interpretation, these phenotypes are most useful when explicitly aligned with four MDT decisions (32).

Iodine quantification may provide a pragmatic link to the hypoxia-vascular axis. Iodine density and standardized iodine measurements could provide a practical surrogate measure of microvascular delivery at the voxel level, which is constrained in PDAC by stromal expansion, elevated interstitial pressure, and hypoxia. After NAT, improvement or stabilization of iodine patterns at the tumor-vessel interface may support exploration despite limited size change. Persistent low or worsening iodine signatures, particularly when concordant with treatment intolerance or CA 19-9 dynamics, may justify continuing, escalating, or modifying NAT (32,47).

Spectral imaging also potentially offers a complementary and stromal-based readout through equilibrium-derived extracellular volume/fractional extracellular space (ECV/fECS). These measures could operationalize the collagen/ECM compartment that dominates PDAC biology by quantifying the expansion of extracellular space within the treated bed, becoming an imaging analog of “response by replacement”. These metrics are MDT relevant in two practical ways. Stromal dominant phenotypes may explain persistent soft tissue without overcalling invasion. In addition, when integrated with clinical variables, extracellular space measures—particularly their spatial heterogeneity—may contribute to early-recurrence risk stratification and downstream surveillance/adjuvant planning (48). Methodologically, the key point is that energy-resolved images, such as iodine maps and VMIs, could be potentially treated as biologically specific input channels for downstream feature extraction and validation, not simply as additional image sets for qualitative interpretation (49-51).

PCCT

If dual-energy CT expands what we can measure, PCCT primarily improves how reliably we can measure it. By counting individual photons and energy-binning the signal at the detector level, PCCT reduces electronic noise, supports higher intrinsic spatial resolution, and makes spectral information routine rather than optional features. These properties address several post-NAT failure modes in PDAC, potentially distinguishing residual tumor from fibrosis, detecting subtle tumor-vessel interface change, and reducing artifact-related uncertainty in the presence of biliary stents (52-55).

PCCT could be most impactful along the hypoxia-vascular axis, where more stable iodine-based readouts can sharpen spatial iodine phenotypes—not only within the tumor mass but, critically, at the tumor-vessel interface where resectability is judged (52,55-57). A pilot study further suggests that iodine density-based metrics on pancreatic PCCT may help discriminate histopathologic treatment response after NAT in PDAC, providing an early clinical anchor for the proposed role of PCCT in therapy-response assessment (58). Along the stroma-ECM axis, higher resolution, lower noise multi-energy data provide more precise estimation of composition linked surrogates, including iodine distribution heterogeneity, which is conceptually aligned with “response by replacement” after NAT (52-54,56). Improved contrast-to-noise ratio and edge definition can also make perivascular and peritumoral gradients—the compartments most likely to reflect interface inflammation and fibrotic remodeling—more consistent inputs for the peritumoral/habitat and modeling analysis (52,55).

By lowering the technical noise floor and improving spectral fidelity, PCCT can make microenvironment linked interface and iodine phenotypes more reproducible inputs for downstream habitat and radiomics modeling (52,55).

Integration with clinical biomarkers, radiomics, habitat analysis and TME

After NAT, CT quantification is better used as building a coherent, biology aligned measurement layer rather than prioritizing any single “best” metric. The key premise is that the uncertainty after treatment mainly stems from changes in the microenvironment, such as pro-fibrotic remodeling, hypoxia-vascular reprogramming, and immune inflammatory responses. These changes can alter CT imaging appearances even without a significant decrease in tumor volume. When post-NAT CT is biologically equivocal, multimodal inputs may help in selected cases, but must prove incremental decision utility beyond CT plus clinical baselines (14,15,22,37,60).

CA 19-9 provides the most clinically established biochemical complement to post-NAT imaging assessment. Rather than functioning as a substitute for CT, CA 19-9 kinetics can serve as a parallel response axis that helps interpret ambiguous imaging findings. A substantial decline or normalization of CA 19-9 after NAT may associate with improved survival, whereas persistently elevated or rising CA 19-9 may indicate residual aggressive biology even when CT morphology appears stable (42,73). Therefore, a decision-oriented post-NAT framework should interpret quantitative CT trajectories together with CA 19-9 dynamics. Concordant improvement in both imaging and CA 19-9 may support surgical exploration or timely transition to resection, whereas discordant findings should reassess systemic disease risk, treatment tolerance, biliary obstruction, and MDT based uncertainty before major surgery is pursued (44,74). However, CA 19-9 also has major biological and clinical limitations. It may be falsely low or uninformative in Lewis antigen-negative patients and may be falsely elevated in the setting of biliary obstruction, cholangitis, or inflammation (43). In such patients, imaging-based assessment, clinical status, alternative biomarkers, and longitudinal rather than single-point interpretation become especially important.

Multimodal fusion is most defensible when it integrates diffusion MRI and PET metabolic readouts as complementary inputs to CT interface and delivery phenotypes (22). Any added modality should show useful decision value over a CT plus clinical baseline and should tolerate clinical challenges (8,70). This principle is illustrated by integrative models showing that quantitative parameters from dual-energy CT and PET/CT combined with clinical pathological variables may improve preoperative identification of patients at high risk for early recurrence; however, such findings may be regarded as adjacent evidence rather than direct post-NAT validation (61). An effective integration strategy is to move from tumor-only phenotyping to multi-compartment, interface led phenotyping. Intratumoral radiomics is useful, particularly delta radiomics that encodes treatment driven remodeling rather than baseline phenotype alone. However, the clinical bottlenecks in PDAC frequently arise at interfaces, where “viable invasion versus scar” is controversial (14,15,37,60). By reasonable design, perivascular ROIs and peritumoral regions encode reader dependent descriptors to reproducible features that directly link to resectability and R0 probability. Moreover, longitudinal tracking can help to define the surgical window. Even when the size changes are limited, improvement in interface focused metrics may support proceeding to exploration, whereas progression may justify extending or adjusting NAT (6,15,60).

Habitat analysis extends this concept by explicitly modeling intratumoral subregions. Habitat analysis segments the tumor into subregions with distinct phenotypes and tracks longitudinal changes in their composition and spatial arrangement, rather than simplifying the lesion toward a single mean attenuation value or a single texture summary. Conceptually, it approximates imaging of microenvironmental habitats. For example, stromal dominant hypoperfused regions adjacent to vascularized residual tumor mass, or an inflammatory tumor periphery adjacent to a fibrotic traction interface (62,63). This framing links habitat level change to MDT decisions. When morphology is not clear, microenvironmental load and interface driven changes may support NAT stewardship, and persistent microenvironmental phenotypes may contribute to early-relapse risk stratification (64). However, it should be interpreted in conjunction with clinical variables rather than in isolation (63).

Deep learning and hybrid AI models are best used as fusion layers, not replacements for mechanistic thinking. In PDAC studies, deep learning models have been trained to predict survival or early recurrence from baseline imaging. Hybrid pipelines that fuse deep-learning-derived latent features with radiomics and clinical markers often outperform either approach alone, because they capture complementary information spanning tumor architecture, stromal organization, and broader anatomical context (17,65,66). AI can also improve the upstream workflow through (semi-)automatic segmentation of tumor and adjacent vessels, reducing interobserver variability and standardizing downstream radiomics and habitat analysis across centers (69).

For clinical translation, methodological defensibility and decision utility are more essential than model complexity across radiomics, habitat analysis, and deep learning (18). Feature definitions are better to obey standardization initiatives, such as the IBSI. Modeling should be reviewed using bias and reporting frameworks such as PROBAST and TRIPOD-AI, and performance claims should prioritize external validation, calibration, and incremental clinical net benefit such as decision curve analysis over headline AUCs (16,18,59).

Strengths, limitations, and future directions

This review has several strengths. First, it organizes post-NAT quantitative CT not as a list of isolated techniques, but as a decision-oriented measurement framework linked to four MDT decisions. Second, it integrates conventional multiphasic CT, perfusion, energy-resolved imaging, radiomics, and habitat-level analysis within a shared microenvironment-aware framework, thereby emphasizing interpretability, methodological defensibility, and clinical translation rather than stand-alone technical performance.

Several barriers still limit the routine use of quantitative CT in NAT-treated PDAC, despite encouraging research results. First of all, clinical questions should be defined in CT operational terms (e.g., what the model is expected to change at the point of care) (6,15,37). Imaging is rarely used in isolation; it complements CA 19-9 dynamics, treatment tolerance, and comorbidity burden. At the same time, CA 19-9 should not be treated as a universally reliable anchor because nonsecretor status, Lewis antigen negativity, biliary obstruction, cholangitis, and inflammatory conditions can distort its interpretation. This reinforces the need for longitudinal, multimodal, and MDT-based interpretation rather than reliance on any single imaging or biochemical marker (43,75). Feasible targets are concrete CT linked decisions rather than generic “good-versus-poor prognosis” classifiers—for example, estimating tumor-vessel interface resectability and the probability of an R0 resection, guiding NAT stewardship when morphology is equivocal, prioritizing longitudinal trajectories rather than single-time-point snapshots when scheduling a surgical window, stratifying early recurrence risk to individualize surveillance intensity and adjuvant planning. Beyond tumor-centered response assessment, post-NAT CT can also provide treatment-related information relevant to MDT risk assessment. In patients receiving oxaliplatin-based NAT, particularly FOLFIRINOX (folinic acid, fluorouracil, irinotecan, and oxaliplatin), chemotherapy-associated liver injury may appear on CT as hepatic steatosis, reduced liver attenuation, heterogeneous hepatic enhancement, liver surface nodularity, splenic volume change, or other CT surrogates of sinusoidal injury and portal hypertension (76). These findings should not be interpreted as direct markers of pathologic tumor response. Instead, they may serve as an additional MDT risk layer, because hepatic parenchymal injury or CT-assessed steatosis may influence surgical-window timing, treatment tolerance, and perioperative morbidity after pancreaticoduodenectomy (76,77). Therefore, post-NAT CT may briefly acknowledge liver parenchymal status when it is relevant to operative timing and MDT risk assessment.

Methodological robustness needs to be sufficient to justify clinical integration. Many radiomics and AI studies in PDAC are retrospective and single-center, with heterogeneous NAT regimens and, critically, heterogeneous CT protocols. Small differences in phase timing, slice thickness, reconstruction kernels, denoising strength, or metal artifact handling can materially alter feature distributions and undermine cross-center universality. The ROI is also part of the biomarker. Tumor-only segmentation measures one biology, whereas perivascular and peritumoral designs explicitly measure target interface and fat plane processes that clinically influence resectability and recurrence. These designs are especially sensitive to contour governance and quality control. Accordingly, protocol discipline and transparent reporting are better to be treated as requirements for deployment rather than post refinements (15,16,63,67).

For adoption, quantitative outputs need to be interpretable at the point of care. Opaque risk scores generated on a separate research workstation are unlikely to change practice. Tools embedded in picture archiving and communication systems (PACS) or structured reporting systems are more likely to be adopted when they provide interpretable overlays, quantify serial change, and report calibrated probabilities. Importantly, CT-based quantification should be designed to reduce inter-reader variability in borderline cases (18,67,68).

Future work should prioritize decision impact, not only technical performance. A coherent, biology-aware CT measurement layer is more likely to accelerate clinical progress. Conventional multiphasic phenotypes can serve as the baseline, with functional and perfusion readouts adding a mechanistic dimension of vascular delivery when morphology is ambiguous. Energy-resolved CT contributes scalable iodine and composition-sensitive surrogates, and PCCT may lower the technical noise floor while improving spatial fidelity, supporting more reproducible interface gradients and heterogeneity features for radiomics, habitat analysis, and hybrid modeling (30,33,52,78). Quantitative imaging and AI cannot be viewed as replacements for expert judgment, but as additional, standardized ways through which to interpret complex NAT response patterns. Future progress will potentially depend on harmonized CT acquisition, strict external validation with calibration, and tighter linkage to pathology and patient-centered endpoints. The field needs evidence that quantitative CT changes MDT decisions and patient management in prospective and controlled settings (30,52,67,79).

Conclusions

Quantitative CT features provide a practical method to connect post-NAT CT appearances with biological and clinical value in PDAC. In this review, we frame post-NAT response as microenvironmental remodeling including stromal reorganization, hypoxia-vascular alteration, and interface inflammation, and relate these processes to CT phenotypes that inform MDT decisions, particularly tumor-vessel interface resectability, NAT stewardship, surgical-window timing, and early recurrence risk stratification. We collect evidence spanning multiphasic quantitative metrics and delta radiomics, perfusion-derived functional readouts, and energy-resolved CT. In addition, these approaches may improve biological specificity and measurement reproducibility, providing an effective substrate for habitat analysis and hybrid AI integration. Clinical translation will depend on harmonized protocols, useful ROI governance, and external validation with calibration and utility analysis.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the Narrative Review reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0239/rc

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0239/prf

Funding: This study was supported by the Program of the Affiliated Hospital of North Sichuan Medical College (No. 2022JB001) and the Doctoral Start-up Fund of North Sichuan Medical College (No. CBY25-QDA41).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-0239/coif). All authors report that this study was supported by the Program of the Affiliated Hospital of North Sichuan Medical College (No. 2022JB001) and the Doctoral Start-up Fund of North Sichuan Medical College (No. CBY25-QDA41). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Siegel RL, Kratzer TB, Wagle NS, et al. Cancer statistics, 2026. CA Cancer J Clin 2026;76:e70043. [Crossref] [PubMed]
2. Janssen QP, O'Reilly EM, van Eijck CHJ, et al. Neoadjuvant Treatment in Patients With Resectable and Borderline Resectable Pancreatic Cancer. Front Oncol 2020;10:41. [Crossref] [PubMed]
3. Javed AA, Habib A, Mahmud O, et al. Prognostic factors in localized pancreatic ductal adenocarcinoma after neoadjuvant therapy and resection: a systematic review and meta-analysis. J Natl Cancer Inst 2025;117:840-67. [Crossref] [PubMed]
4. Turner KM, Wilson GC, Patel SH, et al. ASO Practice Guidelines Series: Management of Resectable, Borderline Resectable, and Locally Advanced Pancreas Cancer. Ann Surg Oncol 2024;31:1884-97. [Crossref] [PubMed]
5. Kwaśniewska D, Fudalej M, Badowska-Kozakiewicz AM, et al. Neoadjuvant Therapy or Upfront Surgery for Pancreatic Cancer-To Whom, When, and How? Cancers (Basel) 2025;17:2584. [Crossref] [PubMed]
6. Soloff EV, Al-Hawary MM, Desser TS, et al. Imaging Assessment of Pancreatic Cancer Resectability After Neoadjuvant Therapy: AJR Expert Panel Narrative Review. AJR Am J Roentgenol 2022;218:570-81. [Crossref] [PubMed]
7. Navez J, Bouchart C, Lorenzo D, et al. What Should Guide the Performance of Venous Resection During Pancreaticoduodenectomy for Pancreatic Ductal Adenocarcinoma with Venous Contact? Ann Surg Oncol 2021;28:6211-22. [Crossref] [PubMed]
8. Panda A, Garg I, Truty MJ, et al. Borderline Resectable and Locally Advanced Pancreatic Cancer: FDG PET/MRI and CT Tumor Metrics for Assessment of Pathologic Response to Neoadjuvant Therapy and Prediction of Survival. AJR Am J Roentgenol 2021;217:730-40. [Crossref] [PubMed]
9. Martin RCG 2nd, Scheon E, Govindarajan N, et al. pRECIST: Guidelines for Response Criteria for Use in Trials Testing Therapeutics in Pancreatic Cancer. Pancreas 2026;55:e547-56. [Crossref] [PubMed]
10. Liot S, Balas J, Aubert A, et al. Stroma Involvement in Pancreatic Ductal Adenocarcinoma: An Overview Focusing on Extracellular Matrix Proteins. Front Immunol 2021;12:612271. [Crossref] [PubMed]
11. Dias Costa A, Väyrynen SA, Chawla A, et al. Neoadjuvant Chemotherapy Is Associated with Altered Immune Cell Infiltration and an Anti-Tumorigenic Microenvironment in Resected Pancreatic Cancer. Clin Cancer Res 2022;28:5167-79. [Crossref] [PubMed]
12. Gartrell RD, Enzler T, Kim PS, et al. Neoadjuvant chemoradiation alters the immune microenvironment in pancreatic ductal adenocarcinoma. Oncoimmunology 2022;11:2066767. [Crossref] [PubMed]
13. Esteva A, Chou K, Yeung S, et al. Deep learning-enabled medical computer vision. NPJ Digit Med 2021;4:5. [Crossref] [PubMed]
14. Wang K, Karalis JD, Elamir A, et al. Delta Radiomic Features Predict Resection Margin Status and Overall Survival in Neoadjuvant-Treated Pancreatic Cancer Patients. Ann Surg Oncol 2024;31:2608-20. [Crossref] [PubMed]
15. Malcolm JA, Tacey M, Gibbs P, et al. Current state of radiomic research in pancreatic cancer: focusing on study design and reproducibility of findings. Eur Radiol 2023;33:6659-69. [Crossref] [PubMed]
16. Zwanenburg A, Vallières M, Abdalah MA, et al. The Image Biomarker Standardization Initiative: Standardized Quantitative Radiomics for High-Throughput Image-based Phenotyping. Radiology 2020;295:328-38. [Crossref] [PubMed]
17. Preuss K, Thach N, Liang X, et al. Using Quantitative Imaging for Personalized Medicine in Pancreatic Cancer: A Review of Radiomics and Deep Learning Applications. Cancers (Basel) 2022;14:1654. [Crossref] [PubMed]
18. Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 2024;385:e078378. [Crossref] [PubMed]
19. Erstad DJ, Sojoodi M, Taylor MS, et al. Fibrotic Response to Neoadjuvant Therapy Predicts Survival in Pancreatic Cancer and Is Measurable with Collagen-Targeted Molecular MRI. Clin Cancer Res 2020;26:5007-18. [Crossref] [PubMed]
20. Janssen BV, van Roessel S, van Dieren S, et al. Histopathological tumour response scoring in resected pancreatic cancer following neoadjuvant therapy: international interobserver study (ISGPP-1). Br J Surg 2022;110:67-75. [Crossref] [PubMed]
21. Matsuda Y, Ohkubo S, Nakano-Narusawa Y, et al. Objective assessment of tumor regression in post-neoadjuvant therapy resections for pancreatic ductal adenocarcinoma: comparison of multiple tumor regression grading systems. Sci Rep 2020;10:18278. [Crossref] [PubMed]
22. Yoo J, Park SJ, Kim H, et al. Prediction of tumor regression grade and identification of prognostic factors using CT and biological features in patients with pancreatic cancer who underwent surgery after neoadjuvant therapy. Acta Radiol 2026;67:388-98. [Crossref] [PubMed]
23. Helms E, Onate MK, Sherman MH. Fibroblast Heterogeneity in the Pancreatic TME. Cancer Discov 2020;10:648-56. [Crossref] [PubMed]
24. Huang H, Wang Z, Zhang Y, et al. Mesothelial cell-derived antigen-presenting cancer-associated fibroblasts induce expansion of regulatory T cells in pancreatic cancer. Cancer Cell 2022;40:656-673. [Crossref] [PubMed]
25. Madsen CD. Pancreatic cancer is suppressed by fibroblast-derived collagen I. Cancer Cell 2021;39:451-3. [Crossref] [PubMed]
26. Muller M, Haghnejad V, Schaefer M, et al. The Immune Landscape of Human Pancreatic Ductal Carcinoma: Key Players, Clinical Implications, and Challenges. Cancers (Basel) 2022;14:995. [Crossref] [PubMed]
27. Bouchart C, Azurmendi Senar O, Navez J, et al. Favorable histo-molecular remodeling of pancreatic ductal adenocarcinoma after neoadjuvant FOLFIRINOX followed by high-dose stereotactic body radiotherapy. Br J Cancer 2026;134:721-34. [Crossref] [PubMed]
28. Hinzpeter R, Kulanthaivelu R, Kohan A, et al. CT Radiomics and Whole Genome Sequencing in Patients with Pancreatic Ductal Adenocarcinoma: Predictive Radiogenomics Modeling. Cancers (Basel) 2022;14:6224. [Crossref] [PubMed]
29. Skornitzke S, Vats N, Mayer P, et al. Pancreatic CT perfusion: quantitative meta-analysis of disease discrimination, protocol development, and effect of CT parameters. Insights Imaging 2023;14:132. [Crossref] [PubMed]
30. Hamdy A, Ichikawa Y, Toyomasu Y, et al. Perfusion CT to Assess Response to Neoadjuvant Chemotherapy and Radiation Therapy in Pancreatic Ductal Adenocarcinoma: Initial Experience. Radiology 2019;292:628-35. [Crossref] [PubMed]
31. Pacella G, Brunese MC, D'Imperio E, et al. Pancreatic Ductal Adenocarcinoma: Update of CT-Based Radiomics Applications in the Pre-Surgical Prediction of the Risk of Post-Operative Fistula, Resectability Status and Prognosis. J Clin Med 2023;12:7380. [Crossref] [PubMed]
32. Mileto A, Ananthakrishnan L, Morgan DE, et al. Clinical Implementation of Dual-Energy CT for Gastrointestinal Imaging. AJR Am J Roentgenol 2021;217:651-63. [Crossref] [PubMed]
33. Barat M, Greffier J, Si-Mohamed S, et al. CT Imaging of the Pancreas: A Review of Current Developments and Applications. Can Assoc Radiol J 2025;76:477-88. [Crossref] [PubMed]
34. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology: Pancreatic Adenocarcinoma. Version 2.2026. [Accessed June 11, 2026]. Available online: https://www.nccn.org/login?ReturnURL=https://www.nccn.org/professionals/physician_gls/pdf/pancreatic.pdf
35. Al-Hawary MM, Francis IR, Chari ST, et al. Pancreatic ductal adenocarcinoma radiology reporting template: consensus statement of the society of abdominal radiology and the american pancreatic association. Gastroenterology 2014;146:291-304. [Crossref] [PubMed]
36. Lee SS, Kim DW, Lee W, et al. Updates on Imaging Assessment of Pancreatic Cancer for Determining Anatomic and Biologic Resectability. Korean J Radiol 2026;27:634-51. [Crossref] [PubMed]
37. Park S, Jang JK, Byun JH, et al. CT in the prediction of margin-negative resection in pancreatic cancer following neoadjuvant treatment: a systematic review and meta-analysis. Eur Radiol 2021;31:3383-93. [Crossref] [PubMed]
38. Castellana R, Fanni SC, Roncella C, et al. Radiomics and deep learning models for CT pre-operative lymph node staging in pancreatic ductal adenocarcinoma: A systematic review and meta-analysis. Eur J Radiol 2024;176:111510. [Crossref] [PubMed]
39. Previtali C, Sartoris R, Rebours V, et al. Quantitative imaging predicts pancreatic fatty infiltration on routine CT examination. Diagn Interv Imaging 2023;104:359-67. [Crossref] [PubMed]
40. Ambrosetti MC, Ambrosetti A, Bariani M, et al. Quantitative Edge Analysis Can Differentiate Pancreatic Carcinoma from Normal Pancreatic Parenchyma. Diagnostics (Basel) 2024;14:1681. [Crossref] [PubMed]
41. Heiselman JS, Horvat N, El Homsi M, et al. Image registration improves inter-reader agreement of objective response in CT assessment of pancreas adenocarcinoma. Eur J Radiol 2026;199:112776. [Crossref] [PubMed]
42. Newhook TE, Vreeland TJ, Griffin JF, et al. Prognosis Associated With CA19-9 Response Dynamics and Normalization During Neoadjuvant Therapy in Resected Pancreatic Adenocarcinoma. Ann Surg 2023;277:484-90. [Crossref] [PubMed]
43. Hartlapp I, Valta-Seufzer D, Siveke JT, et al. Prognostic and predictive value of CA 19-9 in locally advanced pancreatic cancer treated with multiagent induction chemotherapy: results from a prospective, multicenter phase II trial (NEOLAP-AIO-PAK-0113). ESMO Open 2022;7:100552. [Crossref] [PubMed]
44. Dekker EN, van Klaveren D, Verkolf EMM, et al. Carbohydrate antigen 19-9 (CA19-9) response after induction FOLFIRINOX for locally advanced pancreatic cancer identifies patients who may benefit from surgical exploration: multicentre, observational cohort study. Br J Surg 2025;112:znaf011. [Crossref] [PubMed]
45. García-Figueiras R, Goh VJ, Padhani AR, et al. CT perfusion in oncologic imaging: a useful tool? AJR Am J Roentgenol 2013;200:8-19. [Crossref] [PubMed]
46. Bang JY, Breadner D, Nguyen T, et al. Analyzing the effect of neoadjuvant stereotactic ablative body radiotherapy on pancreatic tumor perfusion using computed tomography perfusion. Front Oncol 2026;16:1677923. [Crossref] [PubMed]
47. Shi HY, Lu ZP, Li MN, et al. Dual-Energy CT Iodine Concentration to Evaluate Postoperative Pancreatic Fistula after Pancreatoduodenectomy. Radiology 2022;304:65-72. [Crossref] [PubMed]
48. Wang ZJ, Zhang TT, An C, et al. Estimation of Fractional Extracellular Space at CT for Predicting Chemotherapy Response and Survival in Pancreatic Ductal Adenocarcinoma. AJR Am J Roentgenol 2020;215:610-6. [Crossref] [PubMed]
49. Schwartz FR, Samei E, Marin D. Exploiting the Potential of Photon-Counting CT in Abdominal Imaging. Invest Radiol 2023;58:488-98. [Crossref] [PubMed]
50. Nehra AK, Rajendran K, Baffour FI, et al. Seeing More with Less: Clinical Benefits of Photon-counting Detector CT. Radiographics 2023;43:e220158. [Crossref] [PubMed]
51. Douek PC, Boccalini S, Oei EHG, et al. Clinical Applications of Photon-counting CT: A Review of Pioneer Studies and a Glimpse into the Future. Radiology 2023;309:e222432. [Crossref] [PubMed]
52. Willemink MJ, Persson M, Pourmorteza A, et al. Photon-counting CT: Technical Principles and Clinical Prospects. Radiology 2018;289:293-312. [Crossref] [PubMed]
53. Willemink MJ, Grist TM. Counting Photons: The Next Era for CT Imaging? Radiology 2022;303:139-40. [Crossref] [PubMed]
54. Alkadhi H, Runge V. The Future Arrived: Photon-Counting Detector CT. Invest Radiol 2023;58:439-40. [Crossref] [PubMed]
55. Flohr T, Schmidt B. Technical Basics and Clinical Benefits of Photon-Counting CT. Invest Radiol 2023;58:441-50. [Crossref] [PubMed]
56. Wildberger JE, Alkadhi H. New Horizons in Vascular Imaging With Photon-Counting Detector CT. Invest Radiol 2023;58:499-504. [Crossref] [PubMed]
57. Si-Mohamed SA, Boccalini S, Villien M, et al. First Experience With a Whole-Body Spectral Photon-Counting CT Clinical Prototype. Invest Radiol 2023;58:459-71. [Crossref] [PubMed]
58. Shou A, Bagga B, Hajdu C, et al. A Pilot Study to Assess Pancreatic Adenocarcinoma Treatment Response With Iodine Density From Photon Counting CT. J Comput Assist Tomogr 2026;50:13-9. [Crossref] [PubMed]
59. Moons KGM, Wolff RF, Riley RD, et al. PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration. Ann Intern Med 2019;170:W1-W33. [Crossref] [PubMed]
60. Park S, Sham JG, Kawamoto S, et al. CT Radiomics-Based Preoperative Survival Prediction in Patients With Pancreatic Ductal Adenocarcinoma. AJR Am J Roentgenol 2021;217:1104-12. [Crossref] [PubMed]
61. Zhang Z, Shi L, Pan N, et al. Early postoperative recurrence prediction in pancreatic ductal adenocarcinoma: a nomogram based on dual-energy computed tomography and (18)F-fluorodeoxyglucose positron emission tomography-computed tomography. Quant Imaging Med Surg 2026;16:82. [Crossref] [PubMed]
62. Gao X, Yang L, She T, et al. Habitat-aware radiomics and adaptive 2.5D deep learning predict treatment response and long-term survival in ESCC patients undergoing neoadjuvant chemoimmunotherapy. Eur J Nucl Med Mol Imaging 2026;53:1651-70. [Crossref] [PubMed]
63. Park JE, Kim D, Kim HS, et al. Quality of science and reporting of radiomics in oncologic studies: room for improvement according to radiomics quality score and TRIPOD statement. Eur Radiol 2020;30:523-36. [Crossref] [PubMed]
64. Jiang Y, Zeng J, Sun R, et al. Preoperative radiomics models for predicting risks of distinct recurrence patterns in pancreatic ductal adenocarcinoma based on contrast enhanced CT. Sci Rep 2026;16:7129. [Crossref] [PubMed]
65. Zhang Y, Lobo-Mueller EM, Karanicolas P, et al. Improving prognostic performance in resectable pancreatic ductal adenocarcinoma using radiomics and deep learning features fusion in CT images. Sci Rep 2021;11:1378. [Crossref] [PubMed]
66. Gu Q, Sun H, Liu P, et al. Multiscale deep learning radiomics for predicting recurrence-free survival in pancreatic cancer: A multicenter study. Radiother Oncol 2025;205:110770. [Crossref] [PubMed]
67. Vasey B, Nagendran M, Campbell B, et al. Reporting guideline for the early stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. BMJ 2022;377:e070904. [Crossref] [PubMed]
68. Tejani AS, Cook TS, Hussain M, et al. Integrating and Adopting AI in the Radiology Workflow: A Primer for Standards and Integrating the Healthcare Enterprise (IHE) Profiles. Radiology 2024;311:e232653. [Crossref] [PubMed]
69. Patel H, Zanos T, Hewitt DB. Deep Learning Applications in Pancreatic Cancer. Cancers (Basel) 2024;16:436. [Crossref] [PubMed]
70. Bilreiro C, Andrade L, Marques RM, et al. Diffusion-weighted imaging for determining response to neoadjuvant therapy in pancreatic cancer: a systematic review and meta-analysis. Eur Radiol 2024;34:3238-48. [Crossref] [PubMed]
71. Evangelista L, Zucchetta P, Moletta L, et al. The role of FDG PET/CT or PET/MRI in assessing response to neoadjuvant therapy for patients with borderline or resectable pancreatic cancer: a systematic literature review. Ann Nucl Med 2021;35:767-76. [Crossref] [PubMed]
72. Hwang S, Kim JH, Park J, et al. Key preoperative findings for predicting vascular invasion and recurrence after conversion surgery after chemotherapy in pancreatic cancer. Acta Radiol 2026;67:166-77. [Crossref] [PubMed]
73. Seelen LWF, Doppenberg D, Stoop TF, et al. Minimum and Optimal CA19-9 Response After Two Months Induction Chemotherapy in Patients With Locally Advanced Pancreatic Cancer: A Nationwide Multicenter Study. Ann Surg 2024;279:832-41. [Crossref] [PubMed]
74. Ahmad MU, Javadi CS, Chang JD, et al. Biochemical, Radiographic, or Pathologic Response to Neoadjuvant Chemotherapy in Resected Pancreatic Cancer: Which is Best? Ann Surg 2026;284:336-43. [Crossref] [PubMed]
75. Zhao B, Cheng Q, Cao H, et al. Dynamic change of serum CA19-9 levels in benign and malignant patients with obstructive jaundice after biliary drainage and new correction formulas. BMC Cancer 2021;21:517. [Crossref] [PubMed]
76. Tessier F, Baudouin M, Codjia T, et al. Neoadjuvant Oxaliplatin and Pancreaticoduodenectomy for Pancreatic Ductal Adenocarcinoma: CT Features of Chemotherapy-Associated Liver Injury and Postoperative Complications. Ann Surg Oncol 2026;33:6787-800. [Crossref] [PubMed]
77. Guarneri G, Palumbo D, Pecorelli N, et al. The Impact of CT-Assessed Liver Steatosis on Postoperative Complications After Pancreaticoduodenectomy for Cancer. Ann Surg Oncol 2022;29:7063-73. [Crossref] [PubMed]
78. Albrecht MH, Vogl TJ, Martin SS, et al. Review of Clinical Applications for Virtual Monoenergetic Dual-Energy CT. Radiology 2019;293:260-71. [Crossref] [PubMed]
79. Garcés Albir M, García García F, Mora Oliver I, et al. 3D-PANC study protocol: Is three-dimensional imaging technology superior to computed tomography for assessing the response to neoadjuvant treatment in patients with borderline or locally advanced pancreatic adenocarcinoma? Cir Esp (Engl Ed) 2026;104:800256. [Crossref] [PubMed]