EARLY DETECTION OF RECURRENT OVARIAN CANCER USING ARTIFICIAL INTELLIGENCE IN COMPUTED TOMOGRAPHY

Abstract

Background: Ovarian cancer is one of the most lethal gynecological malignancies due to its subtle symptomatology and high recurrence rate. Its late-stage detection and nonspecific clinical signs hinder timely diagnosis and effective treatment. Traditional imaging methods, while helpful, often lack precision in early recurrence detection, necessitating advanced diagnostic solutions such as artificial intelligence (AI).

Objective: To evaluate the effectiveness of AI algorithms in the early detection of recurrent ovarian cancer using computed tomography (CT) imaging and compare their diagnostic accuracy with conventional radiological assessments.

Duration: Four Months w.e.f 01-02-2025 to 31-5-2025

Methodology: A cross-sectional study was conducted over four months at Kott Khwaja Saeed Hospital. A total of 181 female patients aged 18 and above with diagnosed ovarian cancer (including high-grade serous carcinoma and other subtypes) were included. Contrast-enhanced CT scans were analyzed using AI algorithms, particularly convolutional neural networks (CNN). Data were statistically evaluated using SPSS version 25, with descriptive and inferential analyses including Chi-square and t-tests, and correlation measures.

Results: AI diagnostic models demonstrated perfect performance metrics, achieving 100% accuracy, precision, recall, F1-score, and AUC-ROC across five diagnostic categories. The most common clinical presentations were pelvic pain (28.7%), postmenopausal bleeding (28.2%), and irregular menstrual cycles (27.1%). CT findings predominantly included complex adnexal masses (29.8%) and multilocular cystic lesions (22.7%). AI-assisted imaging significantly outperformed traditional interpretations in early recurrence detection.

Conclusion: Artificial intelligence shows exceptional promise in the early detection of recurrent ovarian cancer through CT imaging. Its integration into clinical workflows can enhance diagnostic accuracy, enable timely intervention, and potentially improve patient survival and quality of life. Continued research and multi-center trials are recommended to validate and standardize AI applications in oncology.