| Number: | 28ESURABS0112 |
| Type: | Scientific Electronic Presentation |
| Authors: | Veronica Celli, Roberta Ninkova, Michele Guerreri, Innocenza Palaia, Federica Tomao, Paola Pricola, Carlo Catalano, Lucia Manganaro |
| Keywords: | Endometrial adenocarcinoma; Magnetic resonance imaging (MRI); Texture analysis; Radiomic analysis; Radiogenomics. |
Endometrial carcinoma (EC) is the most common gynecologic malignancy in the Western world and the fourth leading cause of cancer death in women [1]. It is a typical neoplasm of the post- menopausal age, presenting in about 90% of cases after the age of 50 years, with a mean age at diagnosis of about 63 years[2]. EC is diagnosed at an early stage (Stage I) in 80% of cases because it causes atypical blood loss, an alarm bell in the postmenopausal patient, and it is correlated with a favorable outcome (5-year survival rates above 95%)[3]. Nevertheless, 15-30% of patients have a relapse of disease, hence identifying preoperatively which patients are most at risk of recurrence may have a great impact on therapeutic management, avoiding over- and under-treatment and prolonging survival [4].
The...
...read more
In our multicentric, retrospective study from October 2015 to August 2022, 70 patients at the first diagnosis of EC were enrolled from two different centers: 55 patients from the Department of Gynecology of our Institute, Umberto I Hospital “Sapienza” University of Rome and 15 patients from European Institute of Oncology (IEO), Milan .
The enrolled patients were retrospectively identified from the database of the department of gynecological oncology and from Picture Archiving and Communication System (PACS) of radiological department in order to fulfill the inclusion criteria: (i) patients with postsurgical histological diagnosis of EC, (ii) patients who has a preoperative staging 3 T MRI with standardized protocol and availability of digital images in DICOM format; (iii) all...
...read more
Several of the models obtained by combining some or all of the five selected features showed performance in predicting low-risk EC preoperatively (AUC > 0.50). Figure 5 shows the results obtained from the best performing model in terms of AUC on the test set. The model that performed the best was based on two features only, the sum of squares and the size zone of non-uniformity, both extracted from T2w images. We found an AUC value of 0.70 for the training set and a slightly higher AUC value of 0.72 for the test set. Even if the best model included T2w-derived features only, also ADC-derived features showed to have meaningful information and satisfying AUC value, nevertheless considering the small pool of patients we...
...read more
In this work we show the potential of using a radiomic-based analysis for tumor risk classification. Since radiomics can provide information regarding high-risk factors from standard preoperative MR images, radiomics based models could be useful in patients with EC, helping with preoperative risk stratification and optimal selection of low-risk patients patients who could avoid lymph node dissection, so avoiding overtreatment in low-risk disease. Furthermore, the correct pre-operative identification of the true risk of aggressiveness of the tumor could allow to give a priority in terms of surgical urgency to high-risk patients, optimizing the logistic burden of the surgery departments. We show that a model based on radiomic features extracted from T2w images can be used to classify between low and...
...read more
Same Authors as...
