Please use this identifier to cite or link to this item: https://observatorio.fm.usp.br/handle/OPI/42734
Full metadata record
DC FieldValueLanguage
dc.contributorSistema FMUSP-HC: Faculdade de Medicina da Universidade de São Paulo (FMUSP) e Hospital das Clínicas da FMUSP-
dc.contributor.authorCARDENAS, Diego Armando Cardona-
dc.contributor.authorFERREIRA JUNIOR, Jose Raniery-
dc.contributor.authorMORENO, Ramon Alfredo-
dc.contributor.authorREBELO, Marina de Fatima de Sa-
dc.contributor.authorKRIEGER, Jose Eduardo-
dc.contributor.authorGUTIERREZ, Marco Antonio-
dc.date.accessioned2021-11-22T20:27:34Z-
dc.date.available2021-11-22T20:27:34Z-
dc.date.issued2021-
dc.identifier.citationMEDICAL IMAGING 2021: BIOMEDICAL APPLICATIONS IN MOLECULAR, STRUCTURAL, AND FUNCTIONAL IMAGING, v.11600, article ID 116001D, 7p, 2021-
dc.identifier.isbn978-1-5106-4030-6-
dc.identifier.issn1605-7422-
dc.identifier.urihttps://observatorio.fm.usp.br/handle/OPI/42734-
dc.description.abstractDual-energy subtraction (DES) is a technique that separates soft tissue from bones in a chest radiograph (CR). As DES requires specialized equipment, we propose an automatic method based on convolutional neural networks (CNNs) to generate virtual soft tissue images. A dataset comprising 35 pairs of CR and its soft-tissue version split in training (28 image pairs) and testing (7 image pairs) sets were used with data augmentation. We tested two types of images: the lung region's cropped image and the segmented lung image. The ribs suppression was treated as a local problem, so each image was divided into 784 patches. The U-Net architecture was used to perform bone suppression. We tested two types of loss functions: mean squared error (L-mse) and L-sm, which combines L-mse with the structural similarity index measure (SSIM). Due to the patches overlapping, it was necessary to interpolate the gray levels on the reconstructed image from the predicted patches. Evaluations were based on SSIM and root mean square error (RMSE) over the reconstructed lung area. The combination that presented the best results used the loss L-sm and the segmented lung image as input to the U-Net (SSIM of 0.858 and RMSE of 0.033). We observed that the U-Net has poor performance when trained with cropped images containing all information from the chest cavity and how the loss using local information can improve CR rib bone suppression. Our results suggest that it is possible removing the rib bones accurately in CR using CNN and a patch-based approach.yeng
dc.description.sponsorshipZerbini Foundation-
dc.description.sponsorshipFoxconn Brazil-
dc.language.isoeng-
dc.publisherSPIE-INT SOC OPTICAL ENGINEERINGeng
dc.relation.ispartofMedical Imaging 2021: Biomedical Applications in Molecular, Structural, and Functional Imaging-
dc.relation.ispartofseriesProgress in Biomedical Optics and Imaging-
dc.rightsrestrictedAccesseng
dc.subjectX-Rayeng
dc.subjectBone Suppressioneng
dc.subjectNeural Networkeng
dc.subjectDeep Learningeng
dc.subjectU-Neteng
dc.titleAutomated radiographic bone suppression with deep convolutional neural networkseng
dc.typeconferenceObjecteng
dc.rights.holderCopyright SPIE-INT SOC OPTICAL ENGINEERINGeng
dc.description.conferencedateFEB 15-19, 2021-
dc.description.conferencelocalELECTR NETWORK-
dc.description.conferencenameConference on Medical Imaging - Biomedical Applications in Molecular, Structural, and Functional Imaging-
dc.identifier.doi10.1117/12.2582210-
dc.subject.wosComputer Science, Artificial Intelligenceeng
dc.subject.wosOpticseng
dc.subject.wosRadiology, Nuclear Medicine & Medical Imagingeng
dc.type.categoryproceedings papereng
dc.type.versionpublishedVersioneng
hcfmusp.description.articlenumber116001D-
hcfmusp.description.volume11600-
hcfmusp.origemWOS-
hcfmusp.origem.idWOS:000671880400043-
hcfmusp.publisher.cityBELLINGHAMeng
hcfmusp.publisher.countryUSAeng
hcfmusp.relation.referenceBobadilla J. C. M., 2017, THESIS STATE U CAMPI THESIS STATE U CAMPIeng
hcfmusp.relation.referenceChen S, 2013, IEEE T BIO-MED ENG, V60, P369, DOI 10.1109/TBME.2012.2226583eng
hcfmusp.relation.referenceChen YY, 2019, IEEE ACCESS, V7, P8346, DOI 10.1109/ACCESS.2018.2890300eng
hcfmusp.relation.referenceFerreira J. R, IN PRESSeng
hcfmusp.relation.referenceFerreira JR, 2020, IEEE ENG MED BIO, P1238, DOI 10.1109/EMBC44109.2020.9176517eng
hcfmusp.relation.referenceGusarev M, 2017, DEEP LEARNING MODELS, P1eng
hcfmusp.relation.referenceHwang EJ, 2019, JAMA NETW OPEN, V2, DOI 10.1001/jamanetworkopen.2019.1095eng
hcfmusp.relation.referenceLitjens G, 2017, MED IMAGE ANAL, V42, P60, DOI 10.1016/j.media.2017.07.005eng
hcfmusp.relation.referenceMandell LA, 2007, CLIN INFECT DIS, V44, pS27, DOI 10.1086/511159eng
hcfmusp.relation.referencePazhitnykh I., 2017, LUNG SEGMENTATION 2Deng
hcfmusp.relation.referenceRonneberger O., 2015, P INT C MED IM COMP, P234eng
hcfmusp.relation.referenceSantos Marcel Koenigkam, 2019, Radiol Bras, V52, P387, DOI 10.1590/0100-3984.2019.0049eng
hcfmusp.relation.referenceWang Z, 2004, IEEE T IMAGE PROCESS, V13, P600, DOI 10.1109/TIP.2003.819861eng
hcfmusp.relation.referenceZhao H., 2015, ARXIV PREPRINT ARXIVeng
hcfmusp.relation.referenceZhou B, 2018, AS C COMP VIS AS C COMP VIS, P298eng
dc.description.indexWoSeng
Appears in Collections:

Comunicações em Eventos - FM/MCP
Departamento de Cardio-Pneumologia - FM/MCP

Comunicações em Eventos - HC/InCor
Instituto do Coração - HC/InCor

Comunicações em Eventos - LIM/13
LIM/13 - Laboratório de Genética e Cardiologia Molecular

Comunicações em Eventos - LIM/65
LIM/65 - Laboratório de Investigação Médica em Bioengenharia


Files in This Item:
File Description SizeFormat 
art_FERREIRA JUNIOR_Automated_radiographic_bone_suppression_with_deep_convolutional_neural_2021.PDF
  Restricted Access
publishedVersion (English)932.87 kBAdobe PDFView/Open Request a copy

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.