Ovarian cancer is the leading cause of gynecological cancer-related mortality globally [1, 2] and ultrasound imaging is one of the main tools used in preoperative diagnosis [3]. The advantages of ultrasound imaging include the absence of ionizing radiation, its widespread availability and portability, its low cost and its ability to provide instantaneous results the operator. Ultrasound imaging continues to be a valuable diagnostic tool in many applications such as Point-of-Care Ultrasound (POCUS), extended Focused Assessment with Sonography for Trauma (eFAST) and others [4, 5, 6]. Accurate characterization of ovarian lesions as benign or malignant is of high clinical relevance [7]. Ultrasound plays a vital role in the diagnosis of ovarian masses and Moro et al. [8] described the clinical and ultrasound features of different subclasses of malignant serous ovarian tumors. Moreover, it can distinguish tubal cancer from ovarian cancer even though the prognosis and treatment are similar for these two cancers [9, 10]. At ultrasound examination, lesion size and morphology are the key sonographic features used to triage women with adnexal masses for expectant or interventional management [11]. In surgically removed adnexal masses with a largest diameter of $\le$2.5 cm, Di Legge A et al. [12] found that small unilocular cysts are usually benign, while small non-unilocular masses, particularly ones with solid components, incur a risk of malignancy and pose a clinical dilemma.

There are two major groups, the International Ovarian Tumor Analysis (IOTA) group and the Society of Radiologists in Ultrasound (SRU), that have published imaging criteria for characterization of adnexal lesions [13]. The IOTA group has developed a number of diagnostic strategies, such as Easy Descriptors, Logistic Regression Models 1 and 2 (LR1 and LR2) [14], Simple Rules [15, 16], Simple Rules risk assessment [17], and Assessment of Different Neoplasias in the adneXa (ADNEX) model [18]. The ADNEX model includes clinical and ultrasound information to calculate the likelihood of malignancy and the likelihood of 4 subclasses of malignant tumors, which is clinically very relevant for guiding patient management [19]. The SRU has published expert evidence-based consensus guidelines with criteria for characterization and management of asymptomatic ovarian and other adnexal cysts [20]. Patel-Lippmann et al. [13] found both methods were sensitive for identifying ovarian malignancies and the IOTA simple rules were slightly more accurate than the SRU guidelines in receiver operating characteristic (ROC) analysis.

At present there is no large, multi-center and standardized database of ovarian ultrasound images for teaching and research in China. Yang et al. [21] created a database on focal liver lesions which ensured the quality of ultrasound images, reduced the difference between radiologists’ diagnoses and provided a large-scale data basis for deep learning analysis. Chen et al. [3] examined 278 patients with at least one adnexal (ovarian, paraovarian or tubal) mass to evaluate the diagnostic accuracy of the ADNEX model, using data from a gynecological oncology center in China. Our precious study examined 620 patients from a single oncology center to validate the diagnostic accuracy of the ADNEX model in the hands of nonexpert ultrasonographers [22]. To our knowledge, the database we propose in this study will be the largest, multicentric, and standardized ultrasound ovarian image data resource in China. The size and variety of the database will make it a valuable source for diagnosis as well as education on ovarian lesions.

Here, we describe the aims and methods of our project, to create a large-scale database on ovarian ultrasound images in China, containing standard normal cases, standard abnormal cases and historical classic cases. The number of patients and multi-center design will make the database a valuable resource for a broad range of studies on ovarian diagnostics. With the database as a platform, the next step will be to implement a prospective and ongoing database for future educations and studies.

The duration of the study is expected to be one and a half years, but it can be extended if delays are encountered. When choosing centers, priority will be given to gynecology and obstetrics hospitals, oncology hospitals or tertiary hospitals, so as to ensure timely case collection. At present there are 23 hospitals from 14 provinces that have started or are preparing to collect cases. The manufacturers and types of ultrasound machines vary in hospitals so the instrument specifications will be provided when uploading images to assist in data analysis. Due to different properties of different ultrasonic machines, each center will determine the optimal parameters to set for each specific machine.

In order to ensure the diversity of the database, each center is required to provide cases including SNC, SAC and HCC. Ovarian lesions can be classified as neoplastic or non-neoplastic while ovarian cysts are the most common non-neoplastic lesions [25]. Non-neoplastic cystic lesions of the ovary can be further divided into functional and non-functional cysts and can be diagnosed by pathological or a combination of clinical and imaging features in clinical practice. So only benign or malignant cases must be confirmed by pathological results as well as the level of serum cancer antigen CA125. Besides, our precious study [22] found that when including CA125 in ADNEX model, the performance in discriminating between stage II-IV ovarian cancer (OC) and stage I OC and between stage II-IV OC ovarian metastasis was improved. According to a tertiary center experience [26], ovarian lesions are not commonly seen in pediatric patients, but there are a few reported cases among different pediatric age groups. To ensure the diversity of cases, we have no age limit, but we estimate that there will be fewer pediatric cases.

When the database is completed, we will calculate the ovarian disease component, analyze the ultrasonic characteristics, set up an ovarian ultrasound image library for study and education, validate existing models, and analyze the images by pattern recognition and other advanced techniques.

In conclusion, we are creating a large-scale database of ovarian ultrasound images and accompanying patient medical history and diagnosis obtained from multiple centers in China. The data will be useful for a broad spectrum of studies, ranging from diagnostic analyses to medical education.

JW and PH contributed to study design, data acquisition, writing and revising the manuscript. LG and YT contributed to data collection and database establishment. QW provided clinical expertise for the study design, arranged the data collection, writing and revising the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

This study was performed with ethics approval (Ethics Committee of Beijing Obstetrics and Gynecology Hospital, Capital Medical University. File no. 2019-KY-042-01). Informed consent was obtained from all individual participants included in the study.

The authors would like to thank all sonographers, nurses and stuff at the each center for establishing the database and making this study possible.

This work was supported by National Health Commission Capacity Building and Continuing Education Center (CSLCCC2019JJSJ001) and Beijing Municipal Administration of Hospitals’ Ascent Plan (DFL20151302).

The authors declare no conflict of interest.