Lung cancer mortality is driven by late detection. Lung cancer remains a leading cause of cancer death in western industrialized countries. More than 90% of lung cancer patients diagnosed in Europe die within five years. However, when the primary tumor is confined to the lung without nodal metastases (stage I), the five-year survival rate after curative surgery exceeds 60% and can reach up to 90% for tumors smaller than 3 cm. This dramatic survival difference makes early detection a critical clinical priority.
Early-stage lung cancer is often asymptomatic. Patients with early lung cancer frequently do not exhibit specific symptoms. Dyspnea, cough, and thoracic pain are nonspecific early signs, while hemoptysis may already indicate advanced disease. The absence of distinct early warning signs makes proactive screening essential for catching the disease when it remains surgically curable.
Three complementary early detection approaches exist. This review compared three non-invasive or minimally invasive detection methods: computed tomography (CT) imaging, sputum cytology (including automated sputum cytometry), and a panel of blood-based tumor markers analyzed using fuzzy logic classification. Each method has distinct sensitivity, specificity, and practical advantages that make them potentially complementary rather than competitive.
CT is highly sensitive but generates many false positives. Across multiple screening studies reviewed, computed tomography consistently achieved sensitivity of around 100% for detecting peripheral lung cancers, with stage I detection rates typically around 80%. However, CT also detected extremely high numbers of non-calcified nodules that turned out to be benign. In one representative study, 2,865 non-calcified solitary pulmonary nodules were identified in 7,956 subjects, yet only 40 proved to be lung cancer.
Calculated specificity was unacceptably low in most series. While some studies reported specificity values of 79-95% based on formal test metrics, the calculated specificity accounting for all non-calcified nodules was far lower, ranging from 1% to 11% in most series. Nodule detection rates of 20-50% across screened populations, with over 90% of detected non-calcified nodules being benign, created an unacceptable false positive burden for clinical use.
CT has important detection gaps by histology and location. CT was effective at detecting peripheral, slowly growing adenocarcinomas and squamous cell carcinomas. However, it systematically missed small cell lung cancers (SCLC) and centrally located tumors, because SCLC grows so rapidly that tumors often developed in the interval between intended CT examinations. This limitation meant CT screening had a significant blind spot for an important and aggressive histological subtype.
Occupational exposure populations created additional challenges. In silicosis and asbestosis patients, CT detected many incidental findings including interstitial fibrosis, pleural thickening, and conglomerated nodular masses (progressive massive fibrosis) that could mimic neoplastic lesions. This dramatically complicated differential diagnosis, limiting the practical value of CT as the sole screening tool in these high-risk occupational populations.
Sputum cytology offers a non-invasive, low-cost screening option. Sputum cytology requires no imaging equipment and can be performed at any medical facility using expectoration. It is non-invasive, cheap, and was historically used to complement chest radiography in lung cancer diagnosis. The technique analyzes shed cells from the airways, making it best suited for central tumors that shed cells into the sputum.
Specificity was consistently high across all reviewed studies. In all reviewed studies, sputum cytology demonstrated high specificity between 88% and 100%. Semi-automated sputum cytometry achieved specificity of 92.7% with sensitivity up to 87.5%. This high specificity makes sputum cytology particularly valuable in large screening programs where minimizing false positives is critical, as a positive result carries high predictive value.
Sensitivity varied widely depending on method and cancer type. Manual sputum cytology sensitivity ranged from 21% to 87% across studies, with most results falling around 40-50%. Semi-automated cytometry improved sensitivity substantially, reaching over 80% in some studies. The method was most effective for squamous cell carcinomas (one study reported 100% sensitivity for squamous cell cancers) but less reliable for adenocarcinomas and peripheral tumors that shed fewer cells into airway secretions.
Sample quality and processing method were critical variables. Diagnostic sensitivity was significantly affected by contamination with blood, saliva, or inflammatory cells, and by specimen handling. Only about two-thirds of collected sputum material was suitable for adequate analysis. The number of slides prepared also mattered substantially, with sensitivity increasing from 82.6% with a single slide to 94.0% with seven to eight slides from the same specimen.
Individual tumor markers are insufficient for screening. Blood-based tumor markers including CYFRA 21-1, CEA, NSE, ProGRP, and SCC have historically been excluded from lung cancer screening guidelines due to low individual sensitivity. However, combining multiple markers could improve performance, and the key challenge was finding a mathematical approach that could optimally integrate overlapping marker signals rather than applying simple threshold cutoffs.
Fuzzy logic replaces binary cutoffs with graded membership functions. Fuzzy logic, first described by Zadeh in 1965, replaces rigid yes/no decisions with graded membership functions that express degrees of certainty. Rather than declaring a CYFRA 21-1 value above 3.3 ng/mL simply as elevated, fuzzy logic assigns a value like 2.8 ng/mL a membership of 0.7 in the normal category and simultaneously 0.3 in the elevated category. This continuous grading captures diagnostic uncertainty that binary cutoffs discard.
IF-THEN rules convert marker values to malignancy probability. The fuzzy classifier applies if-then rules linking each marker level to a malignancy designation. For example: IF CYFRA 21-1 is markedly elevated, THEN the case is malignant. Multiple rules can apply simultaneously to a single value, with the MIN-MAX algorithm combining intermediate results. A defuzzification step using the centre of gravity method then converts the combined fuzzy output into a single malignancy index ranging from 0% to 100%.
The classifier combines five markers into a single output score. Candidate markers included NSE (best single marker for SCLC), CYFRA 21-1 (best single marker for NSCLC), CEA, SCC, and C-reactive protein (CRP), which correlates with tumor size and malignancy across histological types. The best three-marker combination for both SCLC and NSCLC was CYFRA 21-1, NSE, and CRP, selected based on their malignant-benign discrimination in the training data.
Fuzzy logic outperformed both individual markers and logistic regression. At a fixed specificity of 95%, the best single marker CYFRA 21-1 achieved 75% sensitivity for NSCLC. Multiple logistic regression combining CYFRA 21-1, NSE, and CRP raised this to 87%. The fuzzy classifier using the same three markers achieved the highest sensitivity at 92%, with an AUC of 0.98 reflecting exceptional overall diagnostic discrimination.
Fuzzy logic showed its greatest advantage in early-stage detection. For NSCLC specifically, the improvement from using fuzzy logic was most pronounced at early stages. CYFRA 21-1 alone detected only 35% of stage I NSCLC, while the fuzzy classifier detected 75% at the same 95% specificity. For stage IIIa disease, fuzzy improved detection from 58% to 84%. These gains in early-stage sensitivity are precisely where clinical benefit from screening is greatest.
SCLC detection was also substantially improved. For small cell lung cancer, the best single marker NSE detected 65% in limited disease and 75% in extended disease at 95% specificity. The fuzzy classifier raised these figures to 94% for limited disease and 97% for extended disease. This matters because SCLC is the histological subtype that CT screening most consistently misses due to its rapid growth between imaging intervals.
Results were validated in high-risk occupational populations. In separate studies of asbestosis and silicosis patients, the fuzzy classifier achieved positive predictive values of 80% and 77.7% respectively, with negative predictive values of 91.5% and 94.8%. These results confirmed that the classifier maintained strong performance in the occupational high-risk populations where CT screening faced its greatest challenges due to confounding incidental findings.
Each method has complementary strengths and weaknesses. CT provides excellent sensitivity for peripheral, slowly growing tumors but generates massive false positive rates and misses SCLC and central tumors. Sputum cytology provides near-perfect specificity but low sensitivity and is limited to centrally located cancers that shed cells into the airway. The fuzzy tumor marker panel provides blood-based, histology-independent detection with high positive and negative predictive values but requires laboratory infrastructure.
Cost-effectiveness concerns limit standalone CT screening. Despite CT's ability to detect stage I lung cancer that is surgically curable, screening programs using CT alone were not considered cost-effective from the perspective of government payers, particularly for former asbestos workers. The high false positive rate drives unnecessary follow-up procedures, patient anxiety, and healthcare costs that reduce the net benefit of the program.
New imaging technologies risk amplifying overdiagnosis. Modern CT scanners with 1 mm slice resolution detect even smaller lung nodules than earlier-generation equipment. While this improves sensitivity further, it also increases detection of very small indolent nodules, contributing to overdiagnosis and unnecessary treatment of benign diseases. This trend reinforces the need for complementary biomarker-based tests that can help distinguish true malignancy from incidental radiological findings.
No single method meets all screening criteria. High diagnostic accuracy with high positive predictive value, low false positive rate, and high sensitivity are simultaneously required for an effective cancer screening program. This review demonstrated that none of the three methods achieves all these criteria individually, but each addresses gaps left by the others.
A combined multi-modal strategy is recommended. For high-risk populations, the authors proposed combining CT, sputum cytology, and tumor marker panel analysis with fuzzy classification. CT contributes sensitivity for peripheral solid tumors; sputum cytology provides high specificity for central squamous cancers; and the fuzzy tumor marker panel adds blood-based detection of all histological types including SCLC, the subtype most frequently missed by CT screening.
Prospective validation of the combined approach is needed. The review emphasized that the practicability of a sequential combined diagnostic strategy should be validated in a prospective study. Sequential testing strategies, where a positive result on one test triggers the next, could provide cumulative sensitivity benefits while controlling costs and minimizing unnecessary invasive follow-up in truly negative cases.