Breast Cancer Detection by Means of Artificial Neural Networks

Abstract

Breast cancer is a fatal disease causing high mortality in women. Constant efforts are being made for creating more efficient techniques for early and accurate diagnosis. Classical methods require oncologists to examine the breast lesions for detection and classification of various stages of cancer. Such manual attempts are time consuming and inefficient in many cases. Hence, there is a need for efficient methods that diagnoses the cancerous cells without human involvement with high accuracies. In this research, image processing techniques were used to develop imaging biomarkers through mammography analysis and based on artificial intelligence technology aiming to detect breast cancer in early stages to support diagnosis and prioritization of high-risk patients. For automatic classification of breast cancer on mammograms, a generalized regression artificial neural network was trained and tested to separate malignant and benign tumors reaching an accuracy of 95.83%. With the biomarker and trained neural net, a computer-aided diagnosis system is being designed. The results obtained show that generalized regression artificial neural network is a promising and robust system for breast cancer detection. The Laboratorio de Innovacion y Desarrollo Tecnologico en Inteligencia Artificial is seeking collaboration with research groups interested in validating the technology being developed.

Diagnostic Interval: Experiences among Women with Breast Cancer in Malaysia

“…Even when lesions are successfully detected, diagnostic delays can occur due to assessment or management recommendation errors [23]. Therefore, Artificial Intelligent technology is being designed to eliminate the unnecessary waiting time as well as reducing human and technical errors in diagnosing BC [24]. Gandhi et al [25] found that 59% of diagnostic errors were caused by three or more process breakdowns, delaying diagnosis by an average of more than a year.…”

Extraction of Pest Insect Characteristics Present in a Mirasol Pepper (Capsicum annuum L.) Crop by Digital Image Processing

“…For the extraction and registration of the insects, the regions of interest (ROIs) property is applied, which allows such features as eccentricity, area, centroid, solidity, and others to be obtained [42,43].…”

Deep learning-based breast region extraction of mammographic images combining pre-processing methods and semantic segmentation supported by Deeplab v3+

“…As a result, there are always background elements, especially those associated with the pectoral muscle region. Both of these regions may contain prenoise, such as the labels used by radiologists to identify certain features and patient information [12], noise (such as salt and pepper noise, gaussian noise, speckling and Poisson noise) and artifacts that can influence the performance of a CAD system [13]. The areas of poor contrast, high and low intensity and the mix of regular or irregular shapes in the pectoral muscle region and around the labels can also confuse CAD systems [14].…”