Evaluation of PD-L1 Expression in Patients With Non-Small Cell Lung Cancer Using DCE-MRI Quantitative Analysis

Abstract

Purpose: The aim is to evaluate the expression of programmed death ligand 1 (PD-L1) in patients with non-small cell lung cancer (NSCLC) using quantitative perfusion parameters based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI).

Methods: A total of 35 patients with a confirmed diagnosis of NSCLC and sufficient tissue pathology were enrolled in the study. The immunohistochemical (IHC) results were used as the gold standard to determine the thresholds for grouping the patients. The patients were divided into three categories based on their PD-L1 expression: (1) PD-L1-negative (IHC < 1%) and PD-L1-positive (IHC ≥ 1%); (2) PD-L1 weak (IHC < 50%) and strong expression (IHC ≥ 50%); and (3) PD-L1 nonexpression (IHC < 1%), low expression (IHC between 1% and 49%), and high expression (IHC ≥ 50%). DCE-MRI datasets were analyzed to acquire histogram parameters, including mean value, uniformity, skewness, kurtosis, entropy, energy, and quantity, of quantitative perfusion parameters using the extended Tofts model (ETM) and the exchange model (ECM). Subsequently, the parameters were compared between the aforementioned groups.

Results: IHC showed PD-L1 < 1% in 20 cases, PD-L1 (1%-49%) in 14 cases, and PD-L1 ≥ 50% in 14 cases. At a threshold of 50%, statistically significant differences were observed for ETM/K^trans (Q25 and Q50), ETM/K_ep (Q10), and ECM/V_e (Q75 and Q90), with values being higher in the weak PD-L1 expression group. With thresholds of 1% and 50%, the results of the pairwise comparison showed that the ECM/V_e (Q75) value in the low PD-L1 expression group was significantly higher than that in the high PD-L1 expression group.

Conclusion: DCE-MRI quantitative analysis is a valuable tool for the evaluation of PD-L1 expression in NSCLC. It provides a noninvasive method that can be employed as an adjunctive technique for the stratification of PD-L1 expression in patients with NSCLC.

Keywords: histogram; magnetic resonance imaging; non–small cell lung cancer; programmed death ligand‐1; quantitative analysis.

FIGURE 1

(a) An AC patient with high PD‐L1 expression. (b–d) Pseudocolor plots of K ^trans, K _ep, and V _e in that order (all from ETM). (e) The histogram parameters of K ^trans are as follows: mean value = 0.085, skewness = 2.095, kurtosis = 13.475, uniformity = 0.260, energy = 0.016, entropy = 6.117, Q5 = 0.010, Q10 = 0.017, Q25 = 0.036, Q50 = 0.075, Q75 = 0.121, Q90 = 0.163, and Q95 = 0.187. (f) The histogram parameters of K _ep are as follows: mean value = 1.231, skewness = 2.598, kurtosis = 10.449, uniformity = 0.107, energy = 0.015, entropy = 5.435, Q5 = 0.053, Q10 = 0.247, Q25 = 0.570, Q50 = 0.958, Q75 = 1.549, Q90 = 2.414, and Q95 = 3.057. (g) The histogram parameters of V _e are as follows: mean value = 0.129, skewness = 3.856, kurtosis = 14.212, uniformity = −0.515, energy = 0.025, entropy = 5.554, Q5 = 0.014, Q10 = 0.023, Q25 = 0.046, Q50 = 0.088, Q75 = 0.118, Q90 = 0.141, and Q95 = 0.167.

FIGURE 2

Immunohistochemical staining of PD‐L1 (×400). (a) A SCC patient with TPS < 1% (PD‐L1 nonexpression). (b) An AC patient with TPS 1%–49% (low PD‐L1 expression). (c) An AC patient with TPS ≥ 50% (high PD‐L1 expression).

FIGURE 3

Scatter plot of quantitative perfusion histogram parameters for predicted PD‐L1 expression (threshold at 50%). V _e (Q75 and Q90) is from ECM; other parameters are from ETM. The short black horizontal line represents the median.

FIGURE 4

ROC curves for DCE‐MRI perfusion histogram parameters predicting 50% PD‐L1 expression at threshold. ETM/K _ep Q10 (AUC = 0.716, 95% CI 0.539–0.855). ETM/K ^trans Q25 (AUC = 0.736, 95% CI 0.560–0.870), Q50 (AUC = 0.720, 95% CI 0.543–0.858). ECM/V _e Q75 (AUC = 0.748, 95% CI 0.573–0.879), Q90 (AUC = 0.724, 95% CI 0.547–0.861). AUC = 0.740 (95% CI 0.564–0.873) for all perfusion histogram parameters.

FIGURE 5

(a) Scatter plot of ECM/V _e (Q75) in the three classified PD‐L1 expression groups. This parameter was statistically different among the PD‐L1 nonexpression, low expression, and high expression groups (p = 0.044). In the pairwise comparison, a statistically different parameter was found between the PD‐L1 low and high expression groups (p = 0.039), and its value was significantly higher in the PD‐L1 low than in the PD‐L1 high expression group. The short black horizontal line represents the median. (b) ROC curves of ECM/V _e (Q75) predicting PD‐L1 low and high expression, the AUC was 0.800 (95% CI 0.570–0.940), and the sensitivity and specificity were 0.909 and 0.700, respectively.