Computer-aided Radiology for Cancer Detection and Therapy Stratification - Benign or Malignant Ovarian Tumors.
Overview
- Phase
- Not Applicable
- Intervention
- Not specified
- Conditions
- Ovarian Cancer
- Sponsor
- Gynaecologisch Oncologisch Centrum Zuid
- Enrollment
- 600
- Locations
- 5
- Primary Endpoint
- Sensitivity and specificity of CADx algorithm
- Status
- Recruiting
- Last Updated
- 4 years ago
Overview
Brief Summary
In women with an ovarian tumor, it is often unclear whether the tumor is benign or malignant. To differentiate, tumor markers (CA125 and CEA), a transvaginal ultrasound and, depending on the ultrasound image and the CA125 concentration, a CT scan are performed. The quality of radiological imaging in diagnosing abdominal pathology is often not accurate enough, making additional interventions no-dig for proper classification and interpretation of the tumor.
Objective: To improve accuracy for distinguishing benign from malignant disease in patients presenting with an ovarian mass by using a computer aided detection algorithm.
Detailed Description
This research focuses on improving the accuracy of the determination of the nature (benign or malignant) of ovarian tumors by making use of artificial intelligence by creating a CT-scan algorithm. This because a correct preoperative classification of ovarian tumors is essential for appropriate treatment. Existing prediction models often lead to unnecessary referrals to gynecological oncology hospitals, resulting in higher costs and increased stress for the patient. It is therefore important to evaluate other strategies to differentiate between benign and malignant ovarian tumors. Artificial Intelligence (AI) for radiology is currently being developed by the Eindhoven University of Technology (TU/e) and Philips Research Europe and may provide a potential solution to this problem. The currently developed algorithm (CADx), using a support vector machine (SVM), showed within a small population of about 100 patients a sensitivity of 74% and specificity of 74%. These are promising results to train this algorithm even further with more CT-scans images and the addition of clinical variables and even liquid biopsies. Type of study: Retrospective study cohort This is a retrospective analysis on known data in which definitive patients diagnosis has already been established and current analysis will not affect treatment plan. No products for patients are used, only computer aided diagnosis is used on existing radiological imaging, namely CT-scans. This study is linked to two other Dutch trials in which ovarian tumor biomarkers are assessed in order to find out the origin of ovarian tumors preoperatively. The first is the HE4-prediction study, with local protocol ID NL58253.031.16. The second is the OVI-DETECT study, with clinicaltrial.gov number NCT04971421.
Investigators
Jurgen M.J. Piek
MD-PhD
Gynaecologisch Oncologisch Centrum Zuid
Eligibility Criteria
Inclusion Criteria
- •patients with an ovarian tumor of which it is unknown whether it is benign or malignant (Risk of Malignancy Index (RMI) \>200)
- •underwent surgery
- •histological proof of tumor
Exclusion Criteria
- •indefinite pathology report
- •lack of correct description of staging in OR report when applicable
Outcomes
Primary Outcomes
Sensitivity and specificity of CADx algorithm
Time Frame: 3 - 4 years
Percentage of correct determination of malignancy by the Risk of Malignancy Index (RMI) compared to exact determination by CAD assessment in patients with an ovarian tumor
Secondary Outcomes
- Sensitivity and specificity of CADx algorithm with additional variables(3 - 4 years)