Background/Objectives: Circulating tumor cells (CTCs) are important biomarkers for predicting prognosis and evaluating treatment efficacy in cancer. We developed the “CTC-Chip” system based on microfluidics, enabling highly sensitive CTC detection and prognostic assessment in lung cancer and malignant pleural mesothelioma. However, the final identification and enumeration of CTCs require manual intervention, which is time-consuming, prone to human error, and necessitates the involvement of experienced medical professionals. Medical image recognition using machine learning can reduce workload and improve automation. However, CTCs are rare in clinical samples, limiting the training data available to construct a robust CTC image recognition system. In this study, we established a highly accurate artificial intelligence-based CTC recognition system by pre-training convolutional neural networks using images from lung cancer cell lines. Methods: We performed transfer learning of convolutional neural networks. Initially, the models were pre-trained using images obtained from lung cancer cell lines. The model’s accuracy was improved by training with a limited number of clinical CTC images. Results: Transfer learning significantly improved the CTC classification accuracy to an average of 99.51%, compared to 96.96% for a model trained solely on pre-trained cell lines (p < 0.05). This approach showed notable efficacy when clinical training images were limited, achieving statistically significant accuracy improvements with as few as 17 clinical CTC images (p < 0.05). Conclusions: Overall, our findings demonstrate that pre-training with cancer cell lines enables rapid and highly accurate automated CTC recognition even with limited clinical data, significantly enhancing clinical applicability and potential utility across diverse cancer diagnostic workflows.

Detecting rare circulating tumor cells (CTCs), malignant cells of primary site origin, in the bloodstream is difficult. We previously constructed a polymeric microfluidic device with high capture efficiency for lung cancer cell lines using antibodies against the epithelial cell adhesion molecule (EpCAM). In this study, we investigated a method for extracting DNA from single cells captured in a microfluidic device and performed targeted sequencing using the Cancer Hotspot Panel v2. Additionally, we employed a fixation method, which enabled more efficient sequencing of the EpCAM-chip-captured cells than existing fixation methods. We used blood samples obtained from three patients with lung cancer to assess the clinical applicability of our method. Directly sequenced samples revealed better coverage uniformity than samples subjected to whole-genome sequencing. Direct sequencing of cells fixed with preserver fluid and 100% ethanol was performed accurately with high coverage uniformity. Our method demonstrated a sensitivity of 99.4%, specificity of 99.5%, and area under the curve of 1 (allele frequency cut-off, 18%). Our approach suggests that single cells captured in a microfluidic device are sufficient for genetic mutation analysis.

Background/Objectives: Published data on reliable prognostic biomarkers for pulmonary pleomorphic carcinoma (PPC), a rare, aggressive subtype of non-small cell lung cancer, are limited. Programmed death-ligand 1 (PD-L1) has been studied; however, the prognostic impact of other immunoregulatory molecules, such as CD73, CD155, and Ki-67 and their combined expression profiles, remains unclear. Methods: We retrospectively analyzed data of 47 patients who had undergone macroscopic complete resection of PPC between January 2000 and December 2022. Immunohistochemistry for PD-L1, CD73, CD155, and Ki-67 was performed. Cut-off values were determined by receiver operating characteristic analysis for cancer-specific death. The primary endpoint was the association between each biomarker and overall survival (OS) or cancer-specific survival (CSS). The secondary endpoint was the prognostic value of combined biomarker profiles, particularly PD-L1 with CD73, CD155, or Ki-67. Results: At a positivity threshold of 1%, the positivity rates were 70.2% for PD-L1, 89.4% for CD73, 91.5% for CD155, and 83.0% for Ki-67. There were no statistically significant differences in OS or CSS between groups with high expression and those with low expression for any single marker. Secondary endpoint analysis showed that low PD-L1 and low CD73 predicted significantly poorer OS (p = 0.046) and low PD-L1 and low Ki-67 predicted significantly poorer CSS (p = 0.039) compared with other combinations. Conclusions: Primary endpoint analyses showed no statistically significant association between any single marker and prognosis in PPC. In contrast, secondary endpoint findings indicated that concurrent low PD-L1 with either low CD73 or low Ki-67 identified subsets with significantly poorer outcomes. The present findings provide a basis for further investigation of biomarker combinations for prognostic assessment in this rare lung cancer. Because of the limited sample size, these findings are exploratory and require validation in larger independent cohorts.