Andrey Bychkov

Commercially Available Artificial Intelligence Solutions for Gynaecologic Cytology Screening and Their Integration Into Clinical Workflow

ABSTRACT Historically, gynaecologic cytology, particularly cervical screening through Pap smear tests, has been instrumental in early cancer detection, but not without its challenges. These include variability in interpretation and the labour‐intensive nature of manual screening processes. The advent of artificial intelligence (AI) technologies, especially machine learning and deep learning, heralds a new era in cytology, offering enhanced accuracy, consistency, and efficiency. These advancements promise to mitigate traditional limitations by automating routine analyses, aiding early cancer detection, and reducing the workload of laboratory personnel. This review thoroughly examines the current status of commercial AI software in gynaecologic cytology screening. We critically assess the capabilities, performance, and impact of these AI tools in a clinical context. Additionally, the review addresses the integration challenges and potential of AI in clinical practice, including workflow integration, regulatory compliance, and ethical considerations. Through this comprehensive analysis, we aim to provide insights into how AI is reshaping gynaecologic cytology, paving the way for more effective disease management and enhanced patient care in women's health.

Evaluation of general-purpose large language models as diagnostic support tools in cervical cytology

The application of general-purpose large language models (LLMs) in cytopathology remains largely unexplored. This study aims to evaluate the accuracy and consistency of a custom version of ChatGPT-4 (GPT), ChatGPT o3, and Gemini 2.5 Pro as diagnostic support tools for cervical cytology. A total of 200 Papanicolaou-stained cervical cytology images were acquired at 40x magnification, each measuring 384 × 384 pixels. These images consisted of 100 cases classified as negative for intraepithelial lesion or malignancy (NILM) and 100 cases across various abnormal categories: 20 low-grade squamous intraepithelial lesion (LSIL), 20 high-grade squamous intraepithelial lesion (HSIL), 20 squamous cell carcinoma (SCC), 20 adenocarcinoma in situ (AIS), and 20 adenocarcinoma (ADC). Diagnostic accuracy and consistency were evaluated by submitting each image to a GPT, ChatGPT o3, and Gemini 2.5 Pro 5-10 times. When distinguishing normal from abnormal cytology, LLMs showed mean sensitivity between 85.4 % and 100 %, and specificity between 67.2 % and 92.7 %. ChatGPT o3 was more accurate in identifying NILM (mean 89.2 % vs. 67.2 %) but less accurate in detecting LSIL (34 % vs. 85 %), HSIL (6 % vs. 63 %), and ADC (28 % vs. 91 %). Chain-of-thought prompting and submitting multiple images of the same diagnosis to ChatGPT o3 and Gemini 2.5 Pro did not significantly improve accuracy. Both models also performed poorly in identifying cervicovaginal infections. ChatGPT o3 and Gemini 2.5 Pro demonstrated complementary strengths in cervical cytology. Due to their low accuracy and inconsistency in abnormal cytology, general-purpose LLMs are not recommended as diagnostic support tools in cervical cytology.

Basaloid Squamous Cell Carcinoma of the Uterine Cervix: Report of a Case With Molecular Analysis

There is a lack of knowledge about molecular alterations in basaloid squamous cell carcinoma (BSCC) of the uterine cervix. A 72-year-old woman with a history of previous subtotal hysterectomy and current vaginal bleeding was referred to our hospital. Initially, adenoid cystic carcinoma (ACC) was diagnosed upon cervical cytology and biopsy. Chest imaging showed multiple metastatic lesions in both lungs. The surgical specimen showed BSCC with diffuse p16 immunoreactivity and negativity for S-100, c-kit, and neuroendocrine markers. There was a focal minor ACC component, which could have explained the previous cytology and biopsy diagnosis. Next-generation sequencing with two different panels showed coexisting PIK3CA mutation and NTRK2 fusion with 10 additional variants of unknown significance ( ATR, DAXX, FAM123B, JAK1 , KEL, MLL2, NOTCH2 , PALB2, POLD1, POLE ). The MYB gene fusions were not identified. The patient received chemotherapy with TRK inhibitor larotrectinib and carboplatin, which caused shrinkage of metastatic lung nodules. This is the first report of cervical BSCC with extensive molecular workup, which detected multiple genetic events, including targetable ones, which are potentially implicated in the development of a tumor. The accumulation of data and further studies on this tumor are necessary to define its diagnostic criteria and its clinical and biological behavior.