Artificial Neural Network‐Based System for PET Volume Segmentation

• DOI: 10.1155/2010/105610
• Publication Date: 2010-09-26T18:36:24Z
• Authors: Mhd Saeed Sharif; Maysam Abbod; Abbes Amira; Habib Zaidi
• Keywords: R895-920; 610; ddc:616.0757; 004; 3. Good health; info:eu-repo/classification/ddc/616.0757; Medical physics. Medical radiology. Nuclear medicine; 03 medical and health sciences; 0302 clinical medicine; Medical technology; R855-855.5; Research Article

Tumour detection, classification, and quantification in positron emission tomography (PET) imaging at early stage of disease are important issues for clinical diagnosis, assessment response to treatment, radiotherapy planning. Many techniques have been proposed segmenting medical data; however, some the approaches poor performance, large inaccuracy, require substantial computation time analysing volumes. Artificial intelligence (AI) can provide improved accuracy save decent amount time. neural networks (ANNs), as one best AI techniques, capability classify quantify precisely lesions model evaluation a specific problem. This paper presents novel application ANNs wavelet domain PET volume segmentation. ANN performance using different training algorithms both spatial domains with number neurons hidden layer is also presented. The determined according experimental results, which stated Levenberg-Marquardt backpropagation algorithm approach application. intelligent system results compared those obtained conventional including thresholding clustering based approaches. Experimental Monte Carlo simulated phantom data sets volumes nonsmall cell lung cancer patients were utilised validate has demonstrated promising results.