"The machine learning MRI radiomic signature tool could effectively classify early-stage breast cancer patients into groups with different risks of ALNM and DFS, and (may) eventually result in a noninvasive approach to guide future clinical practice," Drs. Herui Yao and Erwei Song of Sun Yat Sen University in Guangzhou told Reuters Health by email.
As reported in JAMA Network Open, the study included 1,214 women (median age, 47) diagnosed with early-stage breast cancer at four hospitals in China between 2007-2019, and randomly divided into development (69.9%) and validation (30.1%) cohorts. All participants underwent preoperative MRI scans and were treated with surgery and sentinel lymph node biopsy or ALN dissection.
The team did a univariate analysis to assess the association between clinical characteristics and ALNM or DFS and used the significant clinical risk factors to develop and validate clinical signatures for ALNM and DFS prediction. Taking into consideration the clinical characteristics and radiomic signature covariates, they developed a clinical-radiomic nomogram that could predict ALNM and DFS for a given individual.
The radiomic signature identified ALNM in the development and validation cohorts with areas under the curve (AUCs) of 0.88 and 0.85, respectively; similarly, the clinical-radiomic nomogram accurately predicted ALNM in both cohorts (AUC, 0.92 and 0.90) based on a least absolute shrinkage and selection operator (LASSO)-logistic regression model.
Three-year DFS also was predicted in both cohorts (AUC, 0.81 and 0.73), and the clinical-radiomic nomogram could discriminate high-risk from low-risk patients in the development (hazard ratio, 0.04) and validation (HR, 0.04) cohorts, based on a random forest-Cox regression model.
Further, the clinical-radiomic nomogram was associated with three-year DFS in both cohorts (AUC, 0.89 and 0.90), and, in a decision curve analysis, displayed better clinical predictive usefulness than the clinical or radiomic signature alone.
Drs. Yao and Song said the team will continue to analyze the association between radiomic features and the tumor microenvironment and explore the mechanisms by which radiomic features predict ALNM and recurrence.
"It may be beneficial to (integrate) radiomic signatures with tumor-microenvironment features such as immune cells, long non-coding RNAs, and types of methylated sites, which have better prediction ability and clinical application value," they noted.
Dr. Nathaniel Braman of Case Western University in Cleveland, author of a related editorial, commented in an email to Reuters Health, "If confirmed, an imaging-based approach for diagnosing the spread of breast cancer to the surrounding lymph nodes could help better guide a surgeon's decision as to whether to remove some or all of the lymph nodes surrounding a tumor."
"It might allow us to assess before surgery whether a patient's lymph nodes are clear of disease, thus avoiding the unnecessary removal of a few for biopsy," he noted. "Perhaps more importantly, it could help detect the 5%-10% of patients whose disease has spread and require more aggressive surgical intervention, but are currently being missed by biopsy."
"The authors' algorithm will next need to be validated with a prospective trial that shows that it is effective on patients who have yet to receive treatment," he said. "It will also be important to ensure that it can perform equally well across patient data from new hospitals and MRI scanners that weren't used in the development of the algorithm."
"Ultimately," he concluded, "the most important question to answer - that can't be determined through a retrospective study design - is whether surgical decisions guided by a radiomics tool results in better prognosis and quality of life for breast cancer patients."
Dr. Deanna J. Attai, Assistant Clinical Professor at UCLA David Geffen School of Medicine, also commented by email. "As the techniques and tools are perfected, it does seem that this shows potential for avoiding surgical axillary staging at least in some patients, which will lower risk of lymphedema and other surgery-related complications."
"The biggest caution is that this was a retrospective study - the authors were very careful to point (this) out and noted that a high-quality prospective validation trial is needed," she said. "They evaluated multiple models and reported on what they termed was the best performing one, so even the authors appear cautious regarding the broad applicability of their findings."
SOURCE: https://bit.ly/37nKQyo and https://bit.ly/3mqHejt JAMA Network Open, online December 8, 2020.
By Marilynn Larkin
Posted on
site created by:
© 2024 Medicom Medical Publishers. All rights reserved. Terms and Conditions | Privacy Policy