An investigational deep learning model requiring one histopathologic slide may be effective at predicting the risk of distant recurrence in patients with endometrial cancer, according to novel findings presented by Fremond et al at the American Association for Cancer Research (AACR) Annual Meeting 2023 (Abstract 5695).
Background
Endometrial cancer is the most common type of uterine cancer. While patients with early-stage uterine cancer have demonstrated a 5-year survival rate of 95%, those who develop a distant recurrence have poorer prognoses.
The risk of distant recurrence may be reduced by adjuvant systemic therapy; therefore, correctly identifying patients at high risk and low risk of distant recurrence may be crucial for recommending personalized adjuvant treatment and avoiding unnecessary morbidity from treatment toxicity.
Current methods of distant recurrence risk stratification—which rely on pathologist assessment of histopathologic images—are limited by significant variability between pathologists. While molecular testing of tumor tissue could be of additional value, this method may come with high costs and the need for complex infrastructure. Additionally, the increasing number of prognostic variables has made it difficult to combine the relevant factors into a single risk score.
“Deep learning is a powerful computer-aided predictive technology that has entered the field of pathology because it can be trained to read complex visual information from [histopathologic] slides after digitization,” explained lead study author Sarah Fremond, MSc, a PhD candidate in the Department of Computational Pathology and Deep learning at the Leiden University Medical Center. “In this study, we aimed to evaluate whether a deep learning model could be trained to predict risk of distant recurrence in patients with endometrial cancer using routine histopathologic slides as a cost-effective input,” she added.
Study Methods and Results
In order to develop the new deep learning model, the researchers compiled data from 1,761 patients with endometrial cancer who had not received prior adjuvant chemotherapy—utilizing long-term follow-up data from patients enrolled in the PORTEC-1, PORTEC-2, and PORTEC-3 randomized clinical trials and patients enrolled in three separate clinical cohorts. One representative histopathologic image was used from each of 1,408 patients to train and optimize the model.
“This means that the model was exposed many times to the histopathologic image and to the information regarding the time to distant recurrence in each patient until the model started to recognize visual features that were predictive of distant recurrence,” Dr. Fremond revealed.
To assess the performance and generalizability of the new deep learning model, it was then tested on a data set of 353 patients whose data were not previously used to train the model. The deep learning model identified 89 of these patients as having a low risk of distant recurrence, 175 of them as having an intermediate risk, and 89 of them as having a high risk. These predictions were found to be consistent with the patients’ outcomes: 3.37% of low-risk patients experienced a distant recurrence, compared with 15.43% of intermediate-risk patients and 36% of high-risk patients, respectively, who experienced distant recurrence.
Conclusions
The researchers noted that the results outperformed pathologist-identified features typically used to assign risk groups—such as tumor type, grade, and molecular class.
“This deep learning model is capable of predicting the risk of distant recurrence for patients with endometrial cancer using one digitized histopathologic slide as a cost-effective input. We are currently working on improving performance by integrating clinical variables that cannot be read in the histopathologic slides,” Dr. Fremond said. “Although additional external validation is needed, the performance of this model serves as an important proof of concept that deep learning models have the potential to optimize clinical care for patients with endometrial cancer,” she concluded.
Disclosure: The research in this study was supported by the Hanarth Foundation. For full disclosures of the study authors, visit aacrjournals.org.
