Tumor microenvironment (TME) plays a pivotal role in immuno-oncology, which investigates the intricate interactions between tumors and the human immune system. Specifically, tumor-infiltrating lymphocytes (TILs) are crucial biomarkers for evaluating the prognosis of breast cancer patients and have the potential to refine immunotherapy precision and accurately identify tumor cells in specific cancer types. In this study, we conducted tissue segmentation and lymphocyte detection tasks to predict TIL scores by employing self-supervised learning (SSL) model-based approaches capable of addressing limited labeling data issues. Our experiments showed a 1.9% improvement in tissue segmentation and a 2% improvement in lymphocyte detection over the ImageNet pre-training model. Using these SSL-based models, we achieved a TIL score of 0.718 with a 4.4% improvement. In particular, when trained with only 10% of the entire dataset, the SwAV pre-trained model exhibited a superior performance over other models. Our work highlights improved tissue segmentation and lymphocyte detection using the SSL model with less labeled data for TIL score prediction.

Screening is critical for prevention and early detection of cervical cancer but it is time-consuming and laborious. Supervised deep convolutional neural networks have been developed to automate pap smear screening and the results are promising. However, the interest in using only normal samples to train deep neural networks has increased owing to the class imbalance problems and high-labeling costs that are both prevalent in healthcare. In this study, we introduce a method to learn explainable deep cervical cell representations for pap smear cytology images based on one-class classification using variational autoencoders. Findings demonstrate that a score can be calculated for cell abnormality without training models with abnormal samples, and we localize abnormality to interpret our results with a novel metric based on absolute difference in cross-entropy in agglomerative clustering. The best model that discriminates squamous cell carcinoma (SCC) from normals gives $0.908\pm 0.003$ area under operating characteristic curve (AUC) and one that discriminates high-grade epithelial lesion (HSIL) $0.920\pm 0.002$ AUC. Compared to other clustering methods, our method enhances the V-measure and yields higher homogeneity scores, which more effectively isolate different abnormality regions, aiding in the interpretation of our results. Evaluation using an external dataset shows that our model can discriminate abnormality without the need for additional training of deep models.

Lymphovascular space invasion (LVSI) is the presence of tumor emboli in the endothelial-lined space at the tumor body’s invasive edge. LVSI is one of three Sedlis criteria components—a prognostic tool for early cervical cancer (CC)—essential for indicating poor prognosis, such as lymph node metastasis, distant metastasis, or shorter survival rate. Despite its clinical significance, an in-depth comprehension of the molecular mechanisms or immune dynamics underlying LVSI in CC remains elusive. Therefore, this study investigated tumor-immune microenvironment (TIME) dynamics of the LVSI-positive group in CC. RNA sequencing included formalin-fixed paraffin-embedded (FFPE) slides from 21 CC patients, and differentially expressed genes (DEGs) were analyzed. Functional analysis and immune deconvolution revealed aberrantly enriched PI3K/Akt pathway activation and a heterogenic immune composition with a low abundance of regulatory T cells (Treg) between LVSI-positive and LVSI-absent groups. These findings improve the comprehension of LSVI TIME and immune mechanisms, benefiting targeted LVSI therapy for CC.

Cervical cancer is a significant health challenge, yet it can be effectively prevented through early detection. Cytology-based screening is critical for identifying cancerous and precancerous lesions; however, the process is labor-intensive and reliant on trained experts to scan through hundreds of thousands of mostly normal cells. To address these challenges, we propose a novel distribution-augmented approach using contrastive self-supervised learning for detecting abnormal squamous cervical cells from cytological images. Our method utilizes color augmentations to enhance the model’s ability to differentiate between normal and high-grade precancerous cells; specifically, high-grade squamous intraepithelial lesions (HSILs) and atypical squamous cells–cannot exclude HSIL (ASC-H). Our model was trained exclusively on normal cervical cell images and achieved high diagnostic accuracy, demonstrating robustness against color distribution shifts. We employed kernel density estimation (KDE) to assess cell type distributions, further facilitating the identification of abnormalities. Our results indicate that our approach improves screening accuracy and reduces the workload for cytopathologists, contributing to more efficient cervical cancer screening programs.