Application of artificial neural networks for diagnosis of breastcancer

摘要

We review four current projects pertaining to artificial neuralnetwork (ANN) models that merge radiologist-extracted findings toperform computer aided diagnosis (CADx) of breast cancer. These projectsare: (1) prediction of breast lesion malignancy using mammographicfindings; (2) classification of malignant lesions as in situ vs.invasive cancer; (3) prediction of breast mass malignancy usingultrasound findings; and (4) the evaluation of CADx models in across-institution study. These projects share in common the use offeedforward error backpropagation ANNs. Inputs to the ANNs are medicalfindings such as mammographic or ultrasound lesion descriptors andpatient history data. The output is the biopsy outcome (benign vs.malignant, or in situ vs. invasive cancer) which is being predicted. AllANNs undergo supervised training using actual patient data. These ANNdecision models may assist in the management of patients with breastlesions, such as by reducing the number of unnecessary surgicalprocedures and their associated cost