Spectral CT Scan Before Treatment May Predict How Lymphoma Patients Respond

For patients diagnosed with lymphoma, the weeks between diagnosis and the first round of chemotherapy are filled with uncertainty. Doctors choose a first-line treatment regimen based on the subtype and stage of the disease, but they cannot say with confidence whether a given patient will achieve a complete response or whether the cancer will resist the drugs and continue to progress. A new study published in BMC Medical Imaging suggests that a relatively new imaging technology, dual-layer spectral detector computed tomography, could help close that gap by revealing, before treatment even begins, which patients are likely to respond poorly.

The research, led by Zhengwu Tan and Jing Wang of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology in Wuhan, China, set out to build a statistical prediction tool known as a nomogram that combines routine clinical information with quantitative measurements from spectral CT scans. The team analyzed baseline data from 91 patients with lymphoma who had undergone spectral CT imaging before starting first-line therapy. Their goal was twofold: to predict which patients would fail to achieve a complete response, and to estimate how the imaging and clinical variables would relate to longer-term outcomes such as progression-free survival and overall survival.

To understand why spectral CT is generating excitement in oncology imaging, it helps to look at how it differs from conventional CT. A standard CT scanner measures how much X-ray radiation is absorbed by tissues at a single effective energy level, producing grayscale images in which a tumor may simply appear as a mass. Dual-layer spectral detector CT, by contrast, separates the X-ray beam into two distinct energy spectra simultaneously, using a detector with two layers of scintillator material. Because the two energy datasets are acquired at exactly the same time and in perfect spatial registration, the scanner can derive material-specific information, most notably the concentration of iodine within tissues, as well as the slope of the spectral attenuation curve across different energy levels.

These quantitative parameters carry physiological meaning. Iodine concentration reflects the degree of blood supply and vascular permeability within a lesion, since iodinated contrast agent accumulates in well-perfused tissue. Tumors with different biological behaviors, including different rates of proliferation and angiogenesis, are expected to take up and wash out contrast in characteristic ways. By measuring iodine concentration in both the arterial phase and the venous phase of contrast enhancement, and by calculating derived metrics such as the normalized iodine concentration and the arterial enhancement fraction based on iodine concentration, the researchers hoped to capture a snapshot of tumor biology that conventional CT cannot provide.

The study team measured these spectral parameters from the largest lesion visible on each patient’s baseline scan, in both arterial and venous phases. Treatment response was then assessed according to the Lugano criteria, the internationally accepted standard for evaluating lymphoma response, which classifies patients into those achieving a complete response and those who do not. The researchers divided their cohort accordingly and used univariable and multivariable logistic regression to identify which clinical and imaging variables independently predicted failure to achieve a complete response.

Three predictors emerged from the multivariable analysis: the presence of B symptoms, the Ki67 proliferation index, and a spectral CT parameter called the arterial enhancement fraction based on iodine concentration, abbreviated AEF IC. B symptoms, which include unexplained fever, drenching night sweats, and significant weight loss, are well-established markers of higher tumor burden and more aggressive disease biology. Ki67 is a pathological measure of how rapidly tumor cells are dividing, obtained from biopsy tissue. The inclusion of AEF IC alongside these two established clinical and pathological factors is the study’s most intriguing finding, because it suggests that the vascular characteristics of the tumor, measured noninvasively before treatment, carry independent predictive information about how the disease will respond to first-line therapy.

From these predictors, the team constructed three models: a clinical model using only clinical variables, a spectral CT model using only imaging parameters, and a combined model integrating both. The combined nomogram achieved an apparent area under the receiver operating characteristic curve of 0.866, with a 95 percent confidence interval of 0.791 to 0.938. An AUC of 0.866 indicates good discriminative ability, meaning the model could distinguish complete responders from non-responders considerably better than chance. The combined model outperformed the clinical model alone, although the difference between the combined model and the spectral CT model alone was not statistically significant. The final combined model was internally validated using 1000 bootstrap resamples, a statistical technique that repeatedly resamples the study population to estimate how the model would perform on new patients from the same source, without repeating the variable selection process each time.

Beyond predicting response, the researchers examined whether their models could stratify patients by risk of disease progression. Using Kaplan-Meier analysis and Cox regression, they evaluated progression-free survival and overall survival, with the median follow-up period estimated using the reverse Kaplan-Meier method. Risk stratification based on the clinical model, the spectral CT model, and the combined model was each associated with progression-free survival, but the combined model showed the strongest association: patients placed in the higher-risk category by the combined model had a hazard ratio of 4.05 for shorter progression-free survival, with a 95 percent confidence interval of 1.85 to 8.88. In practical terms, patients flagged as high risk by the combined nomogram faced roughly four times the hazard of disease progression compared with those in the lower-risk group.

The performance of the models was further assessed using calibration analysis, which tests whether predicted probabilities match observed outcomes across the range of risk, and decision curve analysis, which evaluates the clinical usefulness of a model by weighing the benefits of correct predictions against the harms of false ones. These complementary analyses, together with the ROC analysis, form a fairly rigorous internal evaluation framework, and the authors report that the spectral CT and clinical models showed potential for predicting both treatment response and progression-free survival in lymphoma patients.

The study’s authors are careful to note that further external validation is required before such a nomogram can be adopted in routine clinical practice. The cohort of 91 patients was analyzed retrospectively at a single institution, and internal validation by bootstrap resampling, while valuable, cannot fully substitute for testing the model on independent patient populations at other centers. Nevertheless, the findings point toward a future in which the spectral CT scan a lymphoma patient receives at diagnosis, already part of modern staging workups in many hospitals, could do double duty: not only mapping the extent of disease but also generating quantitative signatures of tumor vascularity that, combined with clinical characteristics such as B symptoms and Ki67, help oncologists anticipate treatment failure early. For patients whose predicted risk is high, such information could in principle support closer monitoring, earlier response assessment, or exploratory treatment intensification, pending the external validation studies that will determine whether this promise holds up in the wider clinical world.

Subject of Research: Predicting lymphoma treatment response and survival using spectral CT imaging parameters and clinical characteristics in nomograms

Article Title: Nomograms based on clinical characteristics and baseline dual-layer spectral detector computed tomography imaging parameters: potential clinical utility in predicting the efficacy of first-line lymphoma treatment

Article References: Tan, Z., Liao, D., Li, M., Gui, S., Wu, H., Luo, K., Qin, Q., & Wang, J. (2026). Nomograms based on clinical characteristics and baseline dual-layer spectral detector computed tomography imaging parameters: potential clinical utility in predicting the efficacy of first-line lymphoma treatment. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02802-5

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02802-5

Keywords: dual-layer spectral detector CT, lymphoma, nomogram, treatment response prediction, iodine concentration, progression-free survival, Ki67, B symptoms, Lugano criteria, logistic regression, survival analysis, cancer imaging