Thursday, December 5 | 8:40 a.m.-8:50 a.m. | R1-SSBR10-5 | Room S406A
Attendees in this session will learn about the performance of an MRI-based fully automated deep-learning system that predicts pathological complete response to neoadjuvant chemotherapy in breast cancer.
Jing Gao, MD, of Binzhou Medical University in Yantai, Shandong, China, will present findings showcasing the performance of the system, which achieved high marks in performance testing.
The researchers used pre-enhanced and peak-enhanced phases of dynamic contrast-enhanced MRI (DCE-MRI) for the system. For the study, they evaluated the software's performance in predicting pathological complete response to neoadjuvant chemotherapy and explored the biological basis underlying the system’s prediction work.
The study included 4,041 women who underwent DCE-MRI before chemotherapy from 10 medical centers and were included in The Cancer Genome Atlas (TCGA). The researchers divided the study population into training, internal testing, external testing, and prospective testing sets.
They developed the fully automated deep-learning system with a U-shaped segmentation network and a Siamese-Vision Transformer to perform tumor segmentation and pathologic complete response prediction sequentially.
The system performed well in predicting treatment response in all training and testing sets. These included area under the curve (AUC) values: training set, 0.853; internal testing, 0.850; external testing, range from 0.907 to 0.834; prospective testing, 0.833.
The team also noted that genetic analysis revealed that high scores assigned by the system were associated with the up-regulation of immune-mediated genes and pathways.
With these results in mind, the researchers wrote that the system could serve as an automatic, noninvasive, and reliable tool for strategizing breast cancer treatment. Attend this session to find out more.
