Data mining is the process of releasing concealed information from a large set of database and it can help researchers gain both narrative and deep insights of exceptional understanding of large biomedical datasets. Data mining can exhibit new biomedical and healthcare knowledge for clinical decision making. Medical assessment is very important but complicated problem that should be performed efficiently and accurately. The goal of this paper is to discuss the research contributions of data mining to solve the complex problem of Medical diagnosis prediction. This paper also reviews the various techniques along with their pros and cons. Among various data mining techniques, evaluation of classification is widely adopted for supporting medical diagnostic decisions.

Melanoma is a cancerous lesion in the pigment-bearing basal layers of the epidermis and is the most deadly form of skin cancer, yet it is also the most treatable, with a cure rate for early-stage melanoma of almost 100%. Therefore, there is a need to develop computer-aided diagnostic systems to facilitate the early detection of melanoma. The first step in these systems is skin lesion segmentation. The next essential step is feature extraction and pattern analysis procedures to make a diagnosis. According to the literature, pigment network or reticular pattern is an important diagnostic parameter for melanoma. We decided to work on this automatic melanoma detection system. In this work a brief study has been carried out about the existing segmentation methods adopted by different scholars and a neural network based method is presented for effective segmentation of the melanoma images. In previous work, various methods i.e. Region refinement, adaptive thresholding (AT), gradient vector flow (GVF),adaptive snake (AS);fuzzy-based split-and-merge algorithm (FBSM), k-nearest neighbor, k-means, log filtering etc. has been proposed for the segmentation of melanoma images but their drawback is improper segmentation and more complexity and calculation time . Our method uses testing and training property of neural networks in which a neural object has been created by training it with normal skin and lesion pixels and then same object is used for segmenting other images having similar intensity values. Experimental results shows better results in terms of accuracy values of final segmentation.

No abstract

No abstract