Peter Zhang
Nov 01, 2024 05:01
A groundbreaking AI model developed by UT Southwestern Medical Center researchers can detect breast cancer metastasis without surgery, using advanced MRI analysis.
In a significant advancement in medical technology, researchers at the University of Texas Southwestern Medical Center have developed a deep learning model that can detect the spread of breast cancer without the need for invasive surgery. This AI-based tool analyzes time-series MRIs and clinical data to identify whether cancer cells have metastasized to nearby lymph nodes, a development that could transform treatment planning for doctors and patients alike, according to NVIDIA.
Reducing Invasive Procedures
Currently, doctors often use sentinel lymph node biopsies (SLNB) to determine if breast cancer has spread to the lymph nodes. This procedure involves injecting dye and a radioactive solution near the tumor to identify sentinel nodes, which are then surgically removed for biopsy. Although effective, SLNB is invasive and carries risks such as anesthesia complications, radiation exposure, and post-surgical pain.
The new AI model, however, presents a noninvasive alternative. Utilizing a custom four-dimensional convolutional neural network (4D CNN), the model was trained on dynamic contrast-enhanced MRI (DCE-MRI) data from 350 women diagnosed with breast cancer that had spread to lymph nodes. It processes data in four dimensions, examining 3D MRI scans over time and integrating clinical variables like age and tumor grade to accurately identify cancerous lymph nodes.
High Accuracy and Future Implications
The AI model has demonstrated an impressive 89% accuracy rate in identifying lymph node metastasis, surpassing traditional imaging methods and radiologist assessments. This could potentially spare breast cancer patients from unnecessary procedures like SLNB and axillary lymph node dissection (ALND), reducing associated risks and healthcare resources.
Dr. Dogan Polat, the study’s lead author, emphasized the model’s focus on data from the primary tumor, minimizing the need for additional imaging. “We aim to decrease the need for additional imaging and reduce the number of invasive procedures for patients,” said Dr. Polat, highlighting the model’s potential to enhance patient outcomes and streamline cancer treatment.
Looking Forward
The researchers plan to deploy the AI model in real-world clinical settings to gather data for further validation and refinement. This step is crucial for assessing its effectiveness across a broader range of clinical scenarios and potentially expanding its application to other cancers.
The use of NVIDIA’s A100 and V100 Tensor Core GPUs was pivotal in building and training the model, as noted by Paniz Karbasi, a study coauthor and NVIDIA Senior HPC Engineer. This collaboration underscores the role of cutting-edge technology in advancing medical research and improving diagnostic accuracy.
Image source: Shutterstock
