ABSTRACT
Twenty-five percent of cervical cancers are classified as endocervical adenocarcinomas (EACs), which comprise a highly heterogeneous group of tumors. A histopathologic risk stratification system known as the Silva pattern system was developed based on morphology. However, accurately classifying such patterns can be challenging. The study objective was to develop a deep learning pipeline (Silva3-AI) that automatically analyzes whole slide image-based histopathologic images and identifies Silva patterns with high accuracy. Initially, a total of 202 patients with EACs and histopathologic slides were obtained from Qilu Hospital of Shandong University for developing and internally testing the Silva3-AI model. Subsequently, an additional 161 patients and slides were collected from seven other medical centers for independent testing. The Silva3-AI model was developed using a vision transformer and recurrent neural network architecture, utilizing multi-magnification patches, and its performance was evaluated based on a class-specific area under the receiver-operating characteristic curve. Silva3-AI achieved a class-specific area under the receiver-operating characteristic curve of 0.947 for Silva A, 0.908 for Silva B, and 0.947 for Silva C on the independent test set. Notably, the performance of Silva3-AI was consistent with that of professional pathologists with 10 years' diagnostic experience. Furthermore, the visualization of prediction heatmaps facilitated the identification of tumor microenvironment heterogeneity, which is known to contribute to variations in Silva patterns.
Subject(s)
Adenocarcinoma , Deep Learning , Uterine Cervical Neoplasms , Female , Humans , Uterine Cervical Neoplasms/pathology , Neural Networks, Computer , ROC Curve , Adenocarcinoma/pathology , Tumor MicroenvironmentABSTRACT
Background: The primary objective of this paper was to assess and analyze the top 100 most cited articles currently cited in studies of fertility-sparing treatments for cervical cancer. Methods: Searching the Web of Science Core Collection database for the top 100 most cited articles on fertility-sparing treatments for cervical cancer, different aspects of the articles were analyzed, including countries, journals, institutions, authors, keywords and topics. Results: The search was conducted up to August 2023, and the number of citations for the top 100 articles ranged from 19 to 212. These articles originated from 28 different countries, with Professor Plante, M. from Canada and Professor Sonoda, Y. from the USA having the highest number of articles, both with 10. Professor Plante, M. was the first author of 9 articles and corresponding author of 9 articles. The Memorial Sloan Kettering Cancer Center in the USA published the most articles (21) and received a total of 258 citations. Gynecologic Oncology published 37 of the top 100 articles, with 524 citations and an average of 14.16 citations per article. Conclusions: The study concludes that the USA has made the most significant contributions to this field based on the number of articles, authors, and institutions. Additionally, keyword clustering and burst analysis revealed the research hotspots and future trends in this area.
ABSTRACT
BACKGROUND: Lymph node metastasis (LNM) significantly impacts the prognosis of individuals diagnosed with cervical cancer, as it is closely linked to disease recurrence and mortality, thereby impacting therapeutic schedule choices for patients. However, accurately predicting LNM prior to treatment remains challenging. Consequently, this study seeks to utilize digital pathological features extracted from histopathological slides of primary cervical cancer patients to preoperatively predict the presence of LNM. METHODS: A deep learning (DL) model was trained using the Vision transformer (ViT) and recurrent neural network (RNN) frameworks to predict LNM. This prediction was based on the analysis of 554 histopathological whole-slide images (WSIs) obtained from Qilu Hospital of Shandong University. To validate the model's performance, an external test was conducted using 336 WSIs from four other hospitals. Additionally, the efficiency of the DL model was evaluated using 190 cervical biopsies WSIs in a prospective set. RESULTS: In the internal test set, our DL model achieved an area under the curve (AUC) of 0.919, with sensitivity and specificity values of 0.923 and 0.905, respectively, and an accuracy (ACC) of 0.909. The performance of the DL model remained strong in the external test set. In the prospective cohort, the AUC was 0.91, and the ACC was 0.895. Additionally, the DL model exhibited higher accuracy compared to imaging examination in the evaluation of LNM. By utilizing the transformer visualization method, we generated a heatmap that illustrates the local pathological features in primary lesions relevant to LNM. CONCLUSION: DL-based image analysis has demonstrated efficiency in predicting LNM in early operable cervical cancer through the utilization of biopsies WSI. This approach has the potential to enhance therapeutic decision-making for patients diagnosed with cervical cancer.