AComputer Aided Diagnosis Systemforthe Detection ofPulmonary Nodules in CT Scans Images

Abstract

Lungcancer (both smallcelland non-smallcell)isthe second most co mmon cancerin both men and women (not countingskin cancer).In men,prostate cancer ismore common,whileinwomen breast cancer ismore common.About14%ofallnewcancersare lungcancers.

The American Cancer Society’sestimatesforlungcancerinthe Unit ed Statesfor 2018are:about234,030newcasesoflungcancer (121,68 0in men and 112,350inwomen) ,about154,050deathsfromlungcan cer (83,550in men and 70,500inwomen).

Use CTscanners tostudydetectedoflungcancerwith alarger nu mber of thinnerslices,resultingin the detectionof more nodules.Thi sincreasein thenumber of imagesperCT examination makesthepr ocessof CT interpretation more time consumingand boringfor the r adiologist.Thiscanlead to decreased detectionsensitivityfor nodules ,apart fromthefactthat the majorityofthescreeningcasesare nor mal, and hence diagnostic readingerrorsmanybehard toavoid.The refore,computerized methodsfor nodule detection toassistthe radiol ogist became important.

Computer-aided diagnosis(CAD) system;First,The CT imagesin D ICOMformatwereread bythesystem,and thelungswere extracted fromthethoraxtominimize the Region OfInterest (ROI),Finally,se veralfeatures(geometricand texture)were extracted, tobeused in t he classificationstage. In classification stagewe used three types(SVM, ANNandKNN). The CADsystemwasable toachieve anaccuracyof SVM92%, AN N87%and KNN86% .