Sarif Kumar Naik

Clump splitting based on detection of dominant points from contours

The presence of clumps in biological cell images may degrade the performance of automated disease detection methods using them. We present techniques to split the clumps based on dominant point detection from contours. The selection of optimal dominant points and split lines is achieved using some logical rules based on geometry of the clumps. The scheme first discards some of the dominant points by processing three successive dominant points and labels the remaining points as split points. The split points are then joined by finding pairs of optimal split points that achieve a good split. Pairs of split points that are opposite to each other are joined first and then the remaining split points are joined by using appropriate heuristics. The performance of the scheme is evaluated using several blood smear images and the results show that the method is capable of handling complex clumps.

Doppler based identification of uterine artery and umbilical artery for monitoring pregnancy

In this paper, we present an algorithm to identify umbilical and uterine arteries from a set of four different maternal and fetal arteries using their Doppler signatures. To distinguish these arteries, we use 132 Doppler signals collected from pregnant women with gestational ages between 24 to 40 weeks. Initially we filter them to remove noise; spectrograms are generated to extract good cycles, which are then analyzed to derive independent features that could uniquely represent an artery. A non-linear classification technique using k-NN (k-nearest neighbor) classifier is further applied to identify umbilical and uterine arteries. The proposed algorithm achieves sensitivity and specificity of above 95% and 97% for identification of uterine artery and above 63% and 80% for umbilical artery.

Hybrid SVM - Random Forest classication system for oral cancer screening using LIF spectra

In this paper, a system for oral cancer screening using Laser Induced Fluorescence(LIF) has been developed. A hybrid approach of classification using Support Vector Machine (SVM) and Random Forest (RF) classifiers is proposed. Performance of the classifier is evaluated using several features types such as Wavelet, DFT, LDFT, ILDFT, DCT, LDCT and Slopes features. The most discriminating features are selected using Recursive Feature Elimination(RFE). Analysis of the problem of subset selection from SVM-RFE ranked list is also performed. The hybrid approach has been compared with stand-alone SVM, SVM-RFE and RF classifiers. The proposed technique improves the performance of the classification system significantly. The novelty of the approach lies in the way the most significant features are exstracted in separate modules to arrive at a decision and how the decision are then fused in an intelligent fashion to arrive at a final classification.

Optical Screening of Oral Cancer: Technology for Emerging Markets

Oral cancer is the sixth most common cancer in the world. It is one of the most prevalent cancers in the developing countries of South Asia accounting for one third of the world burden. Sixty percent of the cancers are advanced by the time they are detected. Two methods of optical spectroscopy for detection of oral cancer have been discussed here. These methods are simple, easy to handle and noninvasive. The evaluation of the data is done automatically using pattern recognition techniques, making the screening subjective.

A new feature selection and classification scheme for screening of oral cancer using laser induced fluorescence

Screening for oral cancer in its early stage is of utmost importance for improving the survival rate of oral cancer patients. The current method of visual examination followed by biopsy of suspected cases is subjective with inter and intra personal variations. With a low ratio of oral-cancer experts to patients compounded by the reluctance of patients to undergo biopsy in rural India. The situation cries out for automatic screening device for oral cancer. In this context, optical spectroscopy based on Laser Induced Fluorescence (LIF) has been shown to be a promising technique in distinguishing between cancerous and benign lesions in the mouth. However, it has been observed that it is very difficult to distinguish pre-malignant spectra from malignant and normal spectra recorded in-vivo. Hence, obtaining the most discriminating features from the spectra becomes important. In this article a new method of feature selection is proposed using mean-shift and Recursive Feature Elimination (RFE) techniques to increase discrimination ability of the feature vectors. Performance of the algorithm is evaluated on a in-vivo recorded LIF data set consisting of spectra from normal, malignant and pre-malignant patients. Sensitivity of above 95% and specificity of above 99% towards malignancy are obtained using the proposed method.