Colorectal cancer (CRC) is the third most common cancer worldwide and the second leading cause of cancer-related death. When colorectal cancer spreads, one of the most important early routes is through nearby lymph nodes, small gland-like structures that are part of the immune system.
Lymph node metastasis (LNM) means cancer cells have traveled from the original tumor into surrounding lymph nodes. This significantly affects prognosis and treatment decisions. Patients whose cancer has spread to lymph nodes have roughly 20% lower five-year survival rates compared to those without lymph node involvement.
Before surgery, doctors need to know whether lymph nodes are involved so they can plan the most effective treatment strategy. The NCCN (National Comprehensive Cancer Network) recommends preoperative therapy for node-positive patients to reduce recurrence and improve survival. Accurate pre-surgical staging is therefore essential.
Standard CT scanning is the most commonly used tool for staging colorectal cancer, but conventional CT assesses lymph nodes mainly based on their size and shape. These criteria are imprecise, particularly for small lymph nodes that may contain cancer cells despite appearing normal in size. New, more quantitative imaging approaches are needed.
Conventional CT evaluates lymph nodes by looking at size (typically a short axis greater than 8-10 mm is considered suspicious), border regularity, and internal texture. However, studies show that standard size-based CT criteria have a sensitivity of only 50 to 70 percent for detecting metastatic lymph nodes, meaning many cancerous nodes are missed.
This limitation is especially problematic for micrometastases, which are tiny clusters of cancer cells within normal-sized lymph nodes. These can significantly impact prognosis but are invisible to conventional imaging approaches.
Dual-energy CT (DECT) is an advanced form of CT that uses two different X-ray energy levels simultaneously to gather additional information about tissue composition. By measuring how different tissues absorb different energy levels, DECT can generate quantitative parameters that go far beyond what conventional CT provides.
Among DECT technologies, dual-layer spectral CT (DLSCT) uses a special detector design that simultaneously captures both high and low energy data at the same time and location, ensuring excellent alignment and reproducibility. This study investigated whether DLSCT-derived quantitative measurements could better identify whether lymph nodes in colorectal cancer patients contain cancer cells.
This was a retrospective study conducted at a single hospital in Guangzhou, China. Sixty-six patients with pathologically confirmed colorectal cancer underwent DLSCT imaging within two weeks before surgery between January 2022 and January 2024. A total of 211 lymph nodes were analyzed: 78 that were confirmed cancerous (metastatic) and 133 that were confirmed non-cancerous (non-metastatic) based on tissue examination after surgery.
After surgery, each lymph node visible on CT was matched to the corresponding surgically removed node, verified by pathological examination. This careful node-by-node matching allowed a direct comparison between imaging measurements and the true cancer status of each individual lymph node.
The DLSCT scan covered from the diaphragm to just below the pelvis, using a contrast agent injected into a vein. Images were captured in three phases: arterial phase (AP), venous phase (VP), and equilibrium phase. Two experienced radiologists independently measured each lymph node using standardized regions of interest on post-processing software.
Several quantitative DLSCT parameters were recorded for each lymph node. These included: iodine concentration (IC), a measure of blood supply to the node; normalized iodine concentration (NIC), which adjusts for individual variation; effective atomic number (Zeff), reflecting the elemental composition of tissue; and lambda HU (slope of the spectral Hounsfield unit curve), which describes how tissue absorbs X-rays across different energy levels.
Metastatic lymph nodes showed significantly lower iodine concentration (IC and NIC) in the arterial phase compared to non-metastatic nodes. This likely reflects reduced blood flow in metastatic nodes, where growing cancer cells compress and disrupt normal vascular structures, creating what researchers describe as perfusion defects.
In the venous phase, metastatic nodes showed significantly higher effective atomic number (Zeff and nZeff). This difference may be explained by changes in elemental tissue composition that occur when cancer cells invade a lymph node, altering the proportion of key elements like hydrogen, carbon, and nitrogen.
The lambda HU values (energy curve slope) were significantly higher in metastatic lymph nodes in both phases. A positive correlation was found between the lambda HU values of metastatic lymph nodes and their matched primary tumors, suggesting that cancerous lymph nodes and primary tumors share similar tissue characteristics, a biologically plausible finding given that the metastatic cells originated from the primary tumor.
In terms of traditional morphological features, metastatic nodes were on average larger (8.5 mm vs. 7.4 mm short-axis diameter), more often had irregular internal texture, and were more likely to have ill-defined borders. However, the differences in size between groups were modest, illustrating why size alone is insufficient for accurate staging.
When morphological features alone (size, border, heterogeneity) were used to distinguish metastatic from non-metastatic lymph nodes, the model achieved an AUC of 0.695, with 66.7% sensitivity and 68.4% specificity. This confirms that conventional shape-based CT assessment has significant limitations.
Among individual DLSCT parameters, venous-phase Zeff and lambda HU each achieved AUCs of 0.808 with sensitivities of 87-90%, already outperforming the morphological model. This highlights the added value of tissue composition data beyond simple shape measurements.
The final multivariate model combining multiple DLSCT parameters (including arterial-phase IC and NIC, venous-phase Zeff and lambda HU) reached an impressive AUC of 0.901, with 82.1% sensitivity and 87.2% specificity. This represents a major improvement over morphology alone and would meaningfully reduce both false positives and false negatives in lymph node staging.
Decision curve analysis confirmed that the combined quantitative model provided greater net clinical benefit than morphological criteria across all clinically reasonable decision thresholds. This means that in practice, using the DLSCT-based model would consistently lead to better treatment decisions than relying on conventional CT features.
The lower iodine concentration seen in metastatic lymph nodes aligns with prior research showing that cancer cell growth within a lymph node disrupts normal blood vessel structure. Animal studies have shown that metastatic nodes lose important growth factors for blood vessel sprouting, and that the pressure from growing tumor masses reduces blood flow into the node.
The higher effective atomic number observed in metastatic nodes reflects fundamental changes in tissue composition at the elemental level. When cancer cells invade, the proportions of hydrogen, carbon, and nitrogen change compared to healthy lymphoid tissue, and trace elements with higher atomic numbers increase, all of which the Zeff parameter captures.
The correlation in spectral curve slope (lambda HU) between metastatic nodes and the primary tumor is explained by the nature of metastasis: the cancer cells in the lymph node originated from the primary tumor and carry similar cellular and vascular properties. This homology shows up in the way both tissues absorb X-rays at different energy levels, a concept supported by similar findings in breast cancer.
These findings collectively suggest that DLSCT can detect subtle physiological and compositional changes in lymph nodes that are invisible to conventional imaging. Crucially, these changes can occur even in nodes that appear normal in size, potentially overcoming the fundamental limitation of size-based CT staging.
One important practical consideration is radiation dose. DLSCT uses a single X-ray source with dual-layer detectors, meaning it acquires spectral data simultaneously without any additional radiation compared to conventional single-energy CT. Some studies even suggest that virtual non-contrast images from DLSCT may allow elimination of an additional plain scan, potentially reducing overall radiation exposure to patients.
From a cost perspective, DLSCT is a retrospective acquisition technology, meaning spectral data is automatically available from every scan without requiring any additional scanning or protocol changes. This makes the approach cost-effective and practical to implement in hospitals that already have DLSCT equipment.
For patients, more accurate pre-surgical lymph node staging could lead to better-tailored treatment plans. Patients correctly identified as node-positive could receive preoperative systemic therapy to reduce recurrence risk, while patients correctly identified as node-negative could avoid unnecessary intensification of treatment and its associated side effects.
The ability to identify cancer in normal-sized lymph nodes (smaller than 10 mm) is particularly valuable, as these are the nodes most likely to be missed with conventional CT. Reducing the number of missed metastatic nodes could meaningfully reduce under-staging and improve long-term outcomes for colorectal cancer patients.
The study has important limitations. As a single-center retrospective design with only 66 patients, the findings need validation in larger multicenter prospective studies. The external generalizability of the results cannot be confirmed without replication across different hospitals, scanner models, and patient populations.
The study did not separately analyze lymph nodes by tumor location (right colon, left colon, rectum), which may be important because metastatic patterns and lymph node biology can vary by segment. Future research should stratify patients by tumor location to refine the model's applicability.
All DLSCT measurements were taken using a manual region of interest approach, which depends on radiologist skill. Integrating radiomics or deep learning to automate these measurements could improve reproducibility and potentially identify additional predictive features beyond those analyzed here.
Despite these limitations, this study provides compelling evidence that quantitative DLSCT parameters offer substantial value in colorectal cancer lymph node staging. Combining these measurements with emerging biomarkers and radiomics approaches represents a promising direction toward highly accurate, non-invasive pre-surgical staging of colorectal cancer.