Quantitative information extracted from MRI scans could help predict response to chemoradiation therapy in rectal cancer, according to new research.
Rectal cancer affects 43,000 people in the U.S. every year. Preoperative chemoradiation therapy, a combination of chemotherapy and radiation therapy, is a standard treatment for patients at high risk of recurrence after surgery. Between 10% and 25% of patients achieve a pathological complete response to treatment, meaning there is no residual tumor left.
Regular surveillance, also known as the watch-and-wait approach, can help patients avoid surgical complications and the placement of a permanent stoma, a surgically created path from the large intestine to the abdomen for waste evacuation. However, predicting which patients are good candidates for the watch-and-wait approach is challenging.
"There is an increasing need for evaluation of tumor response preoperatively," said study co-author Jaeseung Shin, MD, Department of Radiology, Yonsei University College of Medicine in Seoul, in a Thursday session.
Radiomics, a method of extracting a large number of quantitative features from medical images, is a promising tool for evaluating rectal cancer patients. While previous studies have shown the potential of radiomics as an indicator of response to treatment, limited data makes the generalizability of previous models less than optimal.
"Therefore, it is required to validate radiomics classification model with a large-scale cohort," Dr. Shin said.
The Power of Radiomics
For the study, Dr. Shin and colleagues evaluated a radiomics model and compared it with qualitative assessments made by radiologists.
The study group included 898 patients with locally advanced rectal cancer who underwent MRI and surgery after chemoradiation therapy between 2009 and 2018. Out of the 898 subjects, 592 were divided into training cohort and 306 into an independent validation cohort.
A total of 1,132 radiomics features were extracted from the MRI scans. Researchers selected the most useful predictive features to build their predictive models. Three experienced abdominal radiologists independently rated tumor regression grade, a classification of cancer response to preoperative treatment, and compared their results to the diagnostic outcome of the radiomics model.
Among the 898 patients, 189 patients, or 21%, achieved a pathological complete response. The radiomics model achieved a sensitivity for predicting treatment response of 80% —significantly higher than the values of the three readers. The model achieved a specificity of 68%.
"In the large-scale validation, the MRI-based radiomics model showed better classification performance compared with qualitative assessment for diagnosing pathologic complete response in patients with locally advanced rectal cancer after chemoradiation therapy," Dr. Shin said.
To further underscore the power of radiomics, Dr. Shin shared images from a 74-year-old male patient whose MRI results produced a discrepancy between visual assessment and radiomics classification. All the radiologists reported a tumor regression grade of 3, a grade that is not expected to have a pathological complete response. However, the radiomics score classified the results as a pathological complete response. Pathological examination of the tumor supported the radiomics prediction.
"The model may provide potentially valuable information to select patients for a watch-and-wait approach," Dr. Shin said.
For More Information:
View the RSNA 2020 session, A Large-Scale Temporal Validation of the MRI Radiomics Model for Prediction of Pathological Complete Response After Preoperative Chemoradiotherapy in Locally Advanced Rectal Cancer — SSGI06 at RSNA2020.RSNA.org.
