Endowing a Content-Based Medical Image Retrieval System with Perceptual Similarity Using Ensemble Strategy

Abstract

Content-based medical image retrieval (CBMIR) is a powerful resource to improve differential computer-aided diagnosis. The major problem with CBMIR applications is the semantic gap, a situation in which the system does not follow the users' sense of similarity. This gap can be bridged by the adequate modeling of similarity queries, which ultimately depends on the combination of feature extractor methods and distance functions. In this study, such combinations are referred to as perceptual parameters, as they impact on how images are compared. In a CBMIR, the perceptual parameters must be manually set by the users, which imposes a heavy burden on the specialists; otherwise, the system will follow a predefined sense of similarity. This paper presents a novel approach to endow a CBMIR with a proper sense of similarity, in which the system defines the perceptual parameter depending on the query element. The method employs ensemble strategy, where an extreme learning machine acts as a meta-learner and identifies the most suitable perceptual parameter according to a given query image. This parameter defines the search space for the similarity query that retrieves the most similar images. An instance-based learning classifier labels the query image following the query result set. As the concept implementation, we integrated the approach into a mammogram CBMIR. For each query image, the resulting tool provided a complete second opinion, including lesion class, system certainty degree, and set of most similar images. Extensive experiments on a large mammogram dataset showed that our proposal achieved a hit ratio up to 10% higher than the traditional CBMIR approach without requiring external parameters from the users. Our database-driven solution was also up to 25% faster than content retrieval traditional approaches.

Retrieval From and Understanding of Large-Scale Multi-modal Medical Datasets: A Review

“…Scale of database Studies small general jpegs: [22]; microscopy: [23]. 2D medium mammograms: [24]; x-rays: [25], [26]; microscopy: [27]- [29]; mixed: [30]. Uni-modal large mammograms: [31], [32]; x-rays: [33]- [36].…”

“…Several retrieval studies have employed medium-to largescale subsets of uni-modal/cue image repositories that are available on the web. Among the found references [24], [31], [32] exploited mammogram ROIs for breast cancer detection from the Digital Database for Screening Mammography (DDSM) repository 9 ; [26], [33]- [36] utilized X-rays from the IRMA repository; [25] used X-rays from the second National Health and Nutrition Examination Survey (NHANES-II) repository 10 to diagnose vertebrae irregularity; and [45] 3 https://ganymed.imib.rwth-aachen.de/irma/ 4 https://imaging.nci.nih.gov/ncia/ 5 http://adni.loni.usc.edu/ 6 http://www.oasis-brains.org/ 7 http://www.imageclef.org/ 8 https://promise12.grand-challenge.org/ 9 http://marathon.csee.usf.edu/Mammography/Database.html 10 https://www.cdc.gov/nchs/nhanes/nhanesii.htm employed CTs from the Lung Image Database Consortium (LIDC) collection 11 for pulmonary nodule retrieval.…”

Arcabouço para Classificação, Recuperação por Conteúdo e Radiômica de Imagens Médicas: uma investigação de biomarcadores quantitativos para o câncer de pulmão

“…A CBIR tem sido descrita como uma das ferramentas computacionais mais promissoras, pois ela potencialmente pode auxiliar o processo de decisão clínica, recuperando em grandes bases de dados casos similares já previamente diagnosticados (MA et al, 2017;BEDO et al, 2016;FERREIRA JUNIOR;OLIVEIRA, 2015;BUGATTI et al, 2014). Sistemas CBIR podem ser muito úteis em um ambiente clínico real, pois podem auxiliar o radiologista na tomada de decisão diagnóstica, ou aumentar sua certeza, valendo-se de um modelo de decisão baseada em exemplo .…”

“…A importância da CBIR como ferramenta de auxílio computadorizado ao processo clínico de interpretação de imagens e diagnóstico de doenças vem crescendo com o passar dos anos (TRAINA et al, 2017;FERREIRA JUNIOR;BEDO et al, 2016;SILVA et al, 2013;KUMAR et al, 2013). Neste capítulo, foi apresentada a avaliação de um algoritmo de recuperação de imagens baseada em atributos de textura e nitidez de borda de nódulos pulmonares presentes em imagens de TC.…”

“…Uma possibilidade de melhorar isto é considerar uma medida de similaridade psicofísica que inclua os atributos das imagens e o julgamento subjetivo de um grupo de radiologistas (NISHIMURA et al, 2013;MURAMATSU et al, 2008). Outra possibilidade é a inclusão de um módulo de realimentação de relevância para reduzir a lacuna semântica, por meio de interações do usuário com o sistema para ponderar os casos recuperados de acordo com a relevância avaliada pelo próprio especialista (BEDO et al, 2016;BUGATTI et al, 2014;AZEVEDO-MARQUES et al, 2008).…”