Background: Carbohydrate sulfotransferases (CHST) were shown to be involved in carcinogenesis. The aim of the study was to assess the diagnostic value of serum CHST7 concentration in differentiation between lung cancer and non-malignant pulmonary inflammations. Methods: Clinical case-control study involving 125 participants was conducted: the control group containing cases of pneumonia and chronic obstructive pulmonary disease was compared to the lung cancer group composed of primary and metastatic cancers. Serum concentrations of CHST7 and routinely used markers including carcinoembryonic antigen (CEA), cytokeratin fragment 21-1 (CYFRA 21-1) and neuron-specific enolase (NSE) were determined for each participant using immunochemical methods. Statistical association, receiver operating characteristic (ROC) analysis and cross-validation were used for the evaluation of CHST7 either as a standalone biomarker or as a part of a biomarker panel. Results: In comparison to the control group, serum CHST7 was elevated in lung cancer (p < 0.001), but no differences between the overall stages of primary cancers were detected (p = 0.828). The differentiation performance in terms of ROC area under curve (AUC) was 0.848 making CHST7 superior biomarker to the NSE (p = 0.031). In comparison to CEA and CYFRA 21-1, the performance difference were not detected. CHST7 was not correlated to other biomarkers, and its addition to the routine biomarker panel significantly improved the cross-validated accuracy (85.6% vs. 75.2%) and ROC AUC (p = 0.004) of the differentiation using a machine learning approach. Conclusions: Serum CHST7 is a promising biomarker for the differentiation between lung cancer and non-malignant pulmonary inflammations.