2014 · 15 citations · 8 references
MedicineComputer ScienceRadiologyComputer VisionDiagnostic ImagingDiagnostic RadiologyCancer DetectionCancer Early DetectionComputed TomographyComputer-aided DiagnosisEarly Cancer DetectionTumor DetectionClassification AccuracyClassification Accuracy AnalysisClassification Accuracy AssessmentComputer TomographyComputer-aided DetectionRegion Of Interest DetectionThresholdingTomography
Lung cancer is very difficult to detect at its early stage with the pathological tests. In past few decades many CAD systems have been designed for early detection of lung tumor to assist the radiologists. This paper presents a CAD system using Computer Tomography images to detect tumor at early stage. The lung regions in CT scan images are segmented using optimal thresholding. To find the region of interest Region growing method is used. Two types of parameters textural and statistical are calculated and analyzed separately. The feature set is applied to the multilayer feed forward back propagation network. The performance of the back propagation neural network is measured in terms of mean square error(MSE). The classification accuracy of the system is 98% with least mean square error. The system is implemented in MATLAB and processing time for classification is 0.22sec.
8