KEY TAKEAWAYS
- The study aimed to investigate a prediction model combining DWI features, DL, and clinical data to identify MSI in EC.
- Researchers noticed that the predictive model integrating DWI features, DL, and clinical data effectively assesses MSI status in EC.
Endometrial cancer (EC) significantly impacts the female reproductive system, with around 30% of patients exhibiting microsatellite instability (MSI). MSI serves as a crucial prognostic biomarker, as patients with MSI-H often experience better outcomes with immunotherapy compared to others.
Traditional MSI testing methods, such as immunohistochemistry (IHC), polymerase chain reaction, and next-generation sequencing, are labor-intensive, invasive, and costly. Therefore, developing a convenient, cost-effective, and noninvasive approach for MSI detection in EC is essential for improving patient care.
Jing Wang and the team aimed to explore the efficacy of a prediction model based on diffusion-weighted imaging (DWI) features extracted from deep learning (DL) and radiomics combined with clinical parameters and apparent diffusion coefficient (ADC) values to identify MSI in EC.
They performed an inclusive analysis involving a cohort of 116 patients with EC, dividing them into training (n = 81) and test (n = 35) sets. Conventional radiomics features and convolutional neural network-based DL features were extracted from DWI. Random forest (RF) and logistic regression classifiers were employed.
The study established various models—DL, radiomics, clinical, ADC, and combined—using DL features, radiomics features, clinical variables, ADC values, and their combinations. The predictive performance of these models was assessed using the area under the receiver operating characteristic curve (AUC), total integrated discrimination index (IDI), net reclassification index (NRI), calibration curves, and decision curve analysis (DCA).
The optimal predictive model based on an RF classifier, incorporated 4 DL features, three radiomics features, 2 clinical variables, and 1 ADC value. This model demonstrated AUC values of 0.989 (95% CI: 0.935-1.000) in the training set and 0.885 (95% CI: 0.731-0.967) in the test set, showing significant improvement compared with the clinical, DL, radiomics, and ADC models (AUC-training = 0.671, 0.873, 0.833, and 0.814; AUC-test = 0.685, 0.783, 0.708, and 0.713, respectively).
The NRI and IDI analyses indicated that the combined model offered enhanced risk reclassification for MSI status compared to the individual clinical, radiomics, DL, and ADC models. Additionally, the calibration curves and DCA demonstrated good consistency and clinical utility of this model, respectively.
The study concluded that the predictive model based on DWI features extracted from DL and radiomics, combined with clinical parameters and ADC values, could effectively assess the MSI status in EC.
The study is funded by the Key Project of Henan Province Medical Science and Technology (LHGJ20220627, 2018020367), Cultivation Project of Guizhou Province for Young Talents in Science and Education (2012)173, and Youth Cultivation Fund Program of Xinxiang Medical University First Affiliated Hospital (QN-2022-B11).
Source: https://pubmed.ncbi.nlm.nih.gov/39171859/
Wang J, Song P, Zhang M, et al. (2024). “A prediction model based on deep learning and radiomics features of DWI for the assessment of microsatellite instability in endometrial cancer.” Cancer Med. 2024;13(16):e70046. doi:10.1002/cam4.70046
