The Wisconsin breast cancer problem: Diagnosis and TTR/DFS time prognosis using probabilistic and generalised regression information classifiers.

This study addresses the breast cancer diagnosis and prognosis problem by employing two neural network architectures with the Wisconsin diagnostic and prognostic breast cancer (WDBC/WPBC) datasets. A probabilistic approach is dedicated to solve the diagnosis problem, detecting malignancy among cases (instances) as derived from fine needle aspirate (FNA) tests, while the second architecture estimates the time interval that possibly contains the right endpoint of disease-free survival (DFS) of the patient. The accuracy of the neural classifiers reaches nearly 98% for the diagnosis and 93% for the prognosis problem, while the prognostic recurrence predictions were evaluated using survival analysis through the Kaplan-Meier approximation method. Both architectures were compared with other similar approaches. The robustness and real-time response of the proposed classifiers were further tested over the web as a potential integrated web-based decision support system.