Abstract: Purpose: Tumor heterogeneityinmedicalimagingisacurrentresearchtrendduetoitspotentialre- lationship withtumormalignancy.Theaimofthisstudyistoanalyzetheeffectofdynamicrangeand matrix sizechangesontheresultsofdifferentheterogeneitymeasures. Materials andmethods: Four patientsharboringthreeglioblastomasandonemetastasiswereconsidered. Sixteen texturalheterogeneitymeasureswerecomputedforeachpatient,withaconfiguration including co-occurrence matrices(CM)features(localheterogeneity)andrun-lengthmatrices(RLM)features (regional heterogeneity).Thecoefficient ofvariationmeasuredagreementbetweenthetexturalmea- sures intwotypesofexperiments:(i) fixing thematrixsizeandchangingthedynamicrangeand(ii) fixing thedynamicrangeandchangingthematrixsize. Results: None ofthemeasuresconsideredwererobustunderdynamicrangechanges.TheCMEntropy and theRLMhighgray-levelrunemphasis(HGRE)weretheoutstandingtexturalfeaturesduetotheir robustness undermatrixsizechanges.Also,theRLMlowgray-levelrunemphasis(LGRE)providedrobust results whenthedynamicrangeconsideredwassufficiently high(morethan8levels).Allofthere- maining texturalfeatureswerenotrobust. Conclusion: Tumor texturestudiesbasedonimageswithdifferentcharacteristics(e.g.multi-center studies) should first fix thedynamicrangetobeconsidered.Forstudiesinvolvingimagesofdifferent resolutions either(i)onlyrobustmeasuresshouldbeused(inourstudyCMentropy,RLMHGREand/or RLM LGRE)or(ii)imagesshouldberesampledtomatchthoseofthelowestresolutionbeforecomputing the texturalfeatures.

Autoría

Descargar referencia