A newly developed machine learning method could reduce instances of unnecessary surgery for some patients with atypical ductal hyperplasia (ADH).
ADH is a condition where a patient has an increase in the number of cells lining the breast ducts. Although not cancerous, those with the condition have a four- to five-fold increase in the risk of developing breast cancer in the future.
ADH is currently identified using mammography and a follow-up core needle biopsy.
This technique has limited accuracy, so the presence of cancer may be underestimated by 10-45%. Because of this, surgical removal is recommended for all ADH cases found on core needle biopsies to determine if the lesion is cancerous.
This means that 70-80% of women undergo a costly and invasive surgical procedure for a benign (but high-risk) lesion.
How machine learning is reducing unnecessary ADH surgeries
A research team at the Norris Cotton Cancer Center at Dartmouth-Hitchcock has developed a machine learning method that can predict whether ADH will lead to cancer.
This means that patients who are low-risk can decide whether close monitoring and hormonal therapy could be an alternative to surgery.
The machine learning approach can identify 98% of all malignant cases prior to surgery, sparing from surgery 16% of women who otherwise would have undergone an unnecessary operation for a benign lesion.
Lead researcher Dr Saeed Hassanpour said:
“Our model can potentially help patients and clinicians choose an alternative management approach in low-risk cases. In the era of personalised medicine, such models can be desirable for patients who value a shared decision-making approach with the ability to choose between surgical excision for certainty versus surveillance to avoid cost, stress, and potential side effects in women at low risk for upgrade of ADH to cancer.”
The team plans to expand the scope of their model by including other high-risk breast lesions such as lobular neoplasia, papillomas, and radial scars.
They also plan to expand their using large external datasets such as national breast cancer registries.
The research has recently been published in JCO Clinical Cancer Informatics.