Abstract:
Great concern has continued to been shown not only in the alarming rate of
death due to Breast Cancer (BCa) cases, but the methods through which the
pathologist use to diagnose BCa. Recent attempts to capture the Whole Slide
Image (WSI) of Invasive Ductal Carcinoma (IDC) breast biopsy tissue with
Medical Imaging techniques showed positive results. As the size of WSI are
accumulating, attempts to analyze WSI based on Image Recognition
Techniques (IRT) to assist the pathologist in diagnostic tasks is becoming
has become a source of relief to histopathological processes. Since the
victory of the team that use Deep Learning at ImageNet Large Scale Visual
Recognition Competition 2012, DL framework has replaced most of the
IRT. This is also the case for Breast Cancer Histological Images. This study
proposes a Deep Convolutional Neural Network Framework (DCNNF) to
enhance the architecture of the state-of-the-art CNN framework to be able
identify Quantitative Semantic Features by leveraging on the kernel selection
methods, optimization techniques and hy-per-parameters tuning. To evaluate
the performance of the proposed approach, we shall build a DCNN model
base on the proposed DCNNF using the IBM Data Scientist Workbench
(IDSW) to utilize the cloud GPU services provided by IBM. The model will
be evaluated with BHI as benchmark dataset. Comparative analysis between
the proposed and the state-of-the-art approach would done base on
classification accu-racy, specificity, sensitivity, precision, recall, ROC and FScore.
