Computer aided clinical risk assessment on stained and unstained images

Abstract

Carcinoma is the 6th most prevalent malignancy worldwide and is a global healthcare challenge. This type of malignancy originated in epithelial cells or tissue, which is the outer cover of the whole body and the lines of different internal organs. Squamous cell carcinoma (SCC) is the second most common types of epithelial cancer. About 90% of the head and neck cancer, 25% of the overall lung cancers are SCC. Moreover, most of the cervical cancers, which arises from harmful Human Papilloma Virus (HPV) are also of SCC type. Primarily, SCC can be curable if detected at early stages. However, it becomes aggressive if not treated at proper time and may spread to distant organs and tissues.The early detection and routine screening are two significant parameters of any diagnostic model for addressing the cancer associated risk and improving the survival rate.Traditional diagnostic system involves the manual assessment of large amount of pathological data, which is tedious and also lead to diagnostic inaccuracy. The major challenge of present clinical research is to develop a complete and accurate CAD-based model, to detect the minute variation in cellular and tissue features at the inception of cancer, by analyzing large amount of clinical data. Moreover, presently adopted non-invasive or semi-invasive procedures for sample collection and application of different label-free microscopic imaging techniques is emerging as a potential alternative of the traditional invasive biopsy method. Present work hypothesized that, the cancer associated mortality rate can be minimizedin automated way, by i) periodically monitoring the wound healing progress of non-healing wounds having tendency to transform in malignancy and ii) routine monitoringoftheindividuals, habitually exposed to the carcinogenic agents for oral cancer detection.To validate the hypothesis, two separate observations were made on the samples collected from lower limb non-healing wounds and the oral epithelial cells of susceptible individuals. This thesis found