Ovarian tumors are a common disease that threatens women's health. They are insidious in onset, have over ten pathological types, and exhibit diverse biological behaviors, making accurate diagnosis a key factor in clinical decision-making and improving prognosis. Introducing AI technology to establish an auxiliary diagnosis system composed of multi-dimensional clinical data, including medical imaging and tumor markers, will greatly enhance diagnosis efficiency by predicting the pathological types of common ovarian tumors.
Our research group has innovatively developed an AI-based ultrasound intelligent auxiliary diagnosis software for ovarian tumors, which has been clinically validated to be effective. This project will build on this by: (1) utilizing a wealth of multi-center retrospective clinical data to combine ultrasound, MRI images, physiological, pathological, and laboratory data to form the first multi-modal ovarian tumor public dataset supporting AI tasks; (2) using convolutional neural network technology to realize multi-modal image multi-classification intelligent recognition on this dataset based on surgical pathology as the standard, and then fuse features at the level of clinical data with the intelligent recognition model to train and validate an auxiliary diagnosis model for predicting the top ten pathological types of ovarian tumors; (3) applying privacy computing and federated learning methods to conduct multi-center, prospective validation and optimization of the above model, ultimately forming a clinical auxiliary diagnosis system that can predict the pathological types of most ovarian tumors and apply it to clinical practice.
1. Continuous cases admitted for diagnosis of ovarian tumor and preparation for surgical treatment;
2. Complete imaging data (ultrasound or MRI) and tumor marker results within 3 months before surgery;
3. Voluntarily sign informed consent.
1. Patients with non-ovarian origin tumor as surgical pathology;
2. Repetitive cases;
3. Cases receiving radiotherapy and chemotherapy;
4. Recurrent cases; 5. Poor image quality of ovarian lesions;
Single group
An auxiliary diagnostic model for ovarian tumors
Diagnostic testThrough collaboration among gynecological oncology teams from three research centers, the multimodal ovarian tumor auxiliary diagnosis system will undergo multicenter, prospective clinical application, with surgical pathology results used as the gold standard. The diagnostic performance of the multimodal ovarian tumor auxiliary diagnosis system, both overall and across different pathological categories, will be evaluated in the main and subsidiary centers by comparing metrics such as specificity, sensitivity, positive predictive value, negative predictive value, and area under the ROC curve (AUC).
Pathological diagnosis
After the surgical specimen is delivered to the pathology department, the pathologist conducts a gross examination, recording the size, shape, surface characteristics, and cut surface features of the specimen (e.g., the proportion of solid and cystic components of the tumor). The specimen is then processed according to standard protocols to prepare tissue sections, which are subsequently stained. Once staining is complete, the slides are independently examined under a microscope by two pathologists, who evaluate the tumor cell morphology, arrangement, and histological characteristics to determine the pathological nature (benign, borderline, or malignant) and type of the tumor. If there is a discrepancy in the diagnoses, the case is referred to a senior pathologist for review, whose final opinion shall prevail.
AUC （Area Under the ROC Curve）
In this study, the area under the curve (AUC) is used as the primary evaluation metric to quantify the diagnostic performance of the new prediction model. AUC is based on the receiver operating characteristic (ROC) curve and evaluates the model's ability to distinguish between positive and negative cases at various thresholds by comparing the model's predictions with the gold standard (pathological diagnosis). Significance of AUC: The AUC value ranges from 0.5 to 1.0, with higher values indicating stronger overall discriminative ability of the model. Specific interpretations are as follows: AUC = 0.5: The model has no diagnostic capability. 0.7 ≤ AUC \< 0.9: The model demonstrates good diagnostic performance. AUC ≥ 0.9: The model shows excellent diagnostic performance. AUC \< 0.7: The model's diagnostic capability is considered low. By calculating the AUC value of the new prediction model, this study assesses its ability to distinguish positive cases from negative cases, thereby verif
Click on the lead sponsor to view all of their trials.
Beijing Shijitan Hospital, Capital Medical University
Lead sponsor
Beijing Friendship Hospital
Collaborator
Beijing Obstetrics and Gynecology Hospital
Collaborator
This trial is not recruiting at the moment. You can still explore other options:
