Classification of endometrial carcinomas by histologic and morphologic features is not reproducible and imperfectly reflects tumor biology, especially in high-grade tumors. Several research teams have defined immunohistochemical and mutation profiles to aid in distinguishing endometrial cancer subtypes. Molecular data have also been used to further stratify risk categories, using gene expression profiling and copy number analysis to determine the risk of recurrence, even in apparent low-risk disease. But the extensive genomic characterization is not easy to be translated into clinical practice. Accumulating data showed both ultrasound and radiomic have a role in identifying patients at "high-risk". Extracting high-dimensional data from clinical ultrasonographic images, radiomic identifies the underlying pathophysiology of tumor tissue. Here, we plan to correlate cancer imaging features and gene expression, in order to categorize endometrial cancer patients into different classes of risk.
To date, molecular/genomic profiling executed after surgery is the most accurate method to asses patients' risk and the need for further surgical and adjuvant treatments. The adoption of preoperative radiomic analysis will allow to plan surgical and adjuvant treatments for endometrial cancer patients. Radiomic will replace the need for extensive molecular/genomic profiling of endometrial cancer. Radiomic data will be harnessed through quantitative image analysis and leveraged via clinical-decision support systems to improve decision-making processes.
The main objective of our project is to adopt radiomics for classifying endometrial cancer according to their classes of risk. To date, molecular/genomic profiling is the best way to assess high-risk tumors. Molecular/genomic profiling is costly (about 1500-2000 euros, per patient); costs related to radiomic profiling are approximately 100 euros, per patient. Additionally, the adoption of radiomics would save time and other resources that are needed to perform molecular/genomic profiling. Based on our preliminary data, our working hypothesis is that radiomics would replace molecular/genomic profiling, thus allowing to save money and other resources for the health care system. In Italy, more than 7,700 endometrial cancer are diagnosed every year (www.airc.it). Given the adoption of radiomics in less than 20% of all newly diagnosed endometrial cancer, the SSN will save more than 2,300,000 euros per year.
Risk analysis, possible problems and solutions:
Possible pitfalls of the project may be represented by:
- Radiomic features variability: the complexity of radiomic/radiogenomics features related to constitutional variables
represents a restrain for using radiomic signals as cancer indicator. Our project exploits rigorous assessment of several
radiomic features and we will we correct those signals based on constitutional variables.
- The application of radiomic on ultrasonographic images: one of the main drawbacks of ultrasound is operator
dependence, which may result in high levels of intraobserver and interobserver variability. However, transvaginal ultrasound
is the best method to identify and categorize endometrial tumors. One of the aims (aim#3) of our project is to assess the
intraobserver and interobserver variability of radiomic features on ultrasonographic images, thus allowing to reproduce our
results on other clinical settings.