Papillary breast lesions include a wide spectrum of pathologies ranging from benign to malignant. The word papillary originates from finger-like projections, or papules, which are seen when these lesions are projected under a microscope. Papillary breast lesions have an array of radiological features at presentations; hence differentiation between benign and malignant based on imaging features is challenging. Histopathological diagnosis is crucial for the distinction and further management of the lesions. Traditionally, tumor and ductal excision is the treatment of choice for malignant and atypical or benign papilloma with imaging discordance. However, current clinical practice guidance advocates complete surgical excision, even for asymptomatic and purely benign papillomas diagnosed on core needle biopsy, as they are highly associated with atypia and malignant upstage on subsequent surgery. Computer aided diagnosis (CAD) is a non-invasive method of diagnosing medical signals/images using advanced image processing followed by soft computing techniques. In this study, we have developed a non-invasive CAD system for differentiating benign versus malignant papillary breast lesions using bi-dimensional empirical mode decomposition (BEMD) and the discrete cosine transform (DCT) followed by locality sensitive discriminant analysis (LSDA). The developed model is validated using a large collection of ultrasound images of papillary breast lesions, and achieved a maximum performance of 98.63% accuracy. We have also developed a breast papillary index, which may in the future be used as a substitute for the conventional soft computing techniques. The developed model can be utilized as a tool to assist radiologists in their routine clinical practice after validation with a larger database.
|Journal||Journal of Ambient Intelligence and Humanized Computing|
|Publication status||Accepted/In press - 2020|
All Science Journal Classification (ASJC) codes
- Computer Science(all)