Diabetic kidney disease (DKD) is a typical microvascular complication characterized by insidious onset and poor prognosis. Conventional contrast-enhanced ultrasound (CEUS) can visualize the microvasculature; however, the bolus injection method involves a high contrast agent concentration and rapid infusion rate, which tends to produce a "flooding" effect. In contrast, the intravenous drip "flash-replenishment" CEUS technique can stably and accurately evaluate microcirculatory hemodynamic changes in renal tissue of DKD patients, potentially providing an imaging basis for the early identification of microcirculatory impairment in DKD.

To investigate the value of quantitative parameters derived from intravenous drip "flash-replenishment" CEUS in assessing DKD microcirculation, to construct and validate a CEUS parameter-based risk prediction model for DKD, and to evaluate its clinical value in the early diagnosis of DKD.

A prospective study was conducted enrolling 85 patients with type 2 diabetes mellitus (T2DM) who visited the outpatient clinic or were hospitalized at Dongzhimen Hospital, Beijing University of Chinese Medicine, from November 2024 to September 2025. Based on urinary protein levels, patients were categorized into the DM-only group (n=27), early-stage DKD group (n=38), and clinical-stage DKD group (n=20). Healthy subjects were concurrently recruited as the healthy control group (n=13). Baseline data were collected from all participants, followed by conventional ultrasound, Doppler ultrasound, and CEUS examinations. The LASSO regression was used for variable selection. The dataset was divided into a training set and a validation set at a ratio of 7∶3. Multivariate Logistic regression analysis was performed on the training set to develop a nomogram prediction model. The receiver operating characteristic (ROC) curve, Hosmer-Lemeshow (H-L) goodness-of-fit test, and decision curve analysis (DCA) were applied to the training and validation sets to evaluate the model's discriminative ability, calibration, and clinical utility, respectively.

Significant differences were observed among the healthy control, DM-only, early-stage DKD, and clinical-stage DKD groups in age, systolic blood pressure (SBP) (P<0.05). Doppler ultrasound: statistically significant differences were found in diastolic velocity (Vd) and resistive index (RI) of the renal artery, segmental artery, and interlobar artery among the four groups (P<0.05). CEUS: statistically significant differences were found in cortical time to peak (TTP), cortical wash-in rate (WiR), cortical half WiR, cortical mean transit time (mTT), medullary peak intensity (PKI), medullary WiR, and medullary half WiR among the four groups (P<0.05). LASSO regression analysis identified two predictors associated with DKD risk: duration of diabetes and cortical WiR. Multivariate Logistic regression analysis confirmed that duration of diabetes (OR=1.169, 95%CI=1.069-1.279) and cortical WiR (OR=0.694, 95%CI=0.499-0.964) were independent predictors of DKD (P<0.05). The ROC curves of the nomogram model showed an AUC of 0.880 (95%CI=0.790-0.969) in the training set and 0.838 (95%CI=0.678-0.998) in the validation set. The H-L goodness-of-fit test indicated mild calibration deviation in the training set (P=0.044) and good calibration in the validation set (P=0.209); combined with calibration curve metrics (training set Eavg=0.081, validation set Eavg=0.124), the overall model calibration was acceptable. DCA showed that the model achieved net benefit superior to the "treat-all" and "treat-none" strategies across a wide range of threshold probabilities (training set 0.09-0.99, validation set 0.08-0.83), demonstrating good clinical applicability.

This study found that longer duration of diabetes and lower cortical WiR are independent predictors of DKD. A nomogram prediction model incorporating these risk factors was established, demonstrating satisfactory overall performance and confirming the value of CEUS in the early diagnosis of DKD.