TUESDAY, May 28, 2024 (HealthDay News) -- A novel natural language understanding (NLU) pipeline can identify the rates of lymphedema and node positivity among women with estrogen receptor-positive (ER+), clinically node-negative (cN0) breast cancer, according to a study published online May 22 in JCO Clinical Cancer Informatics.
Neil Carleton, from the UPMC Hillman Cancer Center in Pittsburgh, and colleagues developed and applied a novel NLU model to examine rates of pathological node positivity (pN+) and rates of lymphedema to assess whether omission of routine axillary staging could be extended to younger patients with ER+/cN0 disease. Data were included for all patients with early-stage ER+, cN0 breast cancer who had sentinel lymph node biopsy (SLNB).
Using the NLU approach, 925 patients aged 55 years or older were identified. The researchers found that the NLU model yielded strong results for breast and arm lymphedema identification, with sensitivity of 100 percent and specificity of 93 percent. The model accrued 93 percent accuracy per patient. Comparison of the NLU model with the Cancer Registry yielded a similar number of patients and no significant differences in the rates of pN+ across ages and clinical T stages. Equally low rates of SLN positivity were seen across ages for patients with clinical stage T1a, T1b, and T1c disease, with higher rates of total lymphedema than rates of pN+.
"These data suggest that the Choosing Wisely recommendation to omit SLNB might be extended to a younger cohort of patients with ER+/cN0 disease," the authors write.
One author is a cofounder of Realyze Intelligence, which provided technology used in the study.
Abstract/Full Text (subscription or payment may be required)
