Cue Integration for medical image annotation

Tatiana Tommasi; Francesco Orabona; Barbara Caputo

doi:10.1007/978-3-540-85760-0_72

Cue Integration for medical image annotation

Tatiana Tommasi, Francesco Orabona, Barbara Caputo

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

This paper presents the algorithms and results of our participation to the image annotation task of ImageCLEFmed 2007. We proposed a multi-cue approach where images are represented both by global and local descriptors. These cues are combined following two SVM-based strategies. The first algorithm, called Discriminative Accumulation Scheme (DAS), trains an SVM for each feature, and considers as output of each classifier the distance from the separating hyperplane. The final decision is taken on a linear combination of these distances. The second algorithm, that we call Multi Cue Kernel (MCK), uses a new Mercer kernel which can accept as input different features while keeping them separated. The DAS algorithm obtained a score of 29.9, which ranked fifth among all submissions. The MCK algorithm with the one-vs-all and with the one-vs-one multiclass extensions of SVM scored respectively 26.85 and 27.54. These runs ranked first and second among all submissions. © 2008 Springer-Verlag Berlin Heidelberg.

Original language	English (US)
Title of host publication	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer Verlag
Pages	577-584
Number of pages	8
ISBN (Print)	3540857591
DOIs	https://doi.org/10.1007/978-3-540-85760-0_72
State	Published - Jan 1 2008
Externally published	Yes

Bibliographical note

Generated from Scopus record by KAUST IRTS on 2023-09-25

Access to Document

10.1007/978-3-540-85760-0_72

Cite this

@inproceedings{548138f0716241de91b152485321ee15,

title = "Cue Integration for medical image annotation",

abstract = "This paper presents the algorithms and results of our participation to the image annotation task of ImageCLEFmed 2007. We proposed a multi-cue approach where images are represented both by global and local descriptors. These cues are combined following two SVM-based strategies. The first algorithm, called Discriminative Accumulation Scheme (DAS), trains an SVM for each feature, and considers as output of each classifier the distance from the separating hyperplane. The final decision is taken on a linear combination of these distances. The second algorithm, that we call Multi Cue Kernel (MCK), uses a new Mercer kernel which can accept as input different features while keeping them separated. The DAS algorithm obtained a score of 29.9, which ranked fifth among all submissions. The MCK algorithm with the one-vs-all and with the one-vs-one multiclass extensions of SVM scored respectively 26.85 and 27.54. These runs ranked first and second among all submissions. {\textcopyright} 2008 Springer-Verlag Berlin Heidelberg.",

author = "Tatiana Tommasi and Francesco Orabona and Barbara Caputo",

note = "Generated from Scopus record by KAUST IRTS on 2023-09-25",

year = "2008",

month = jan,

day = "1",

doi = "10.1007/978-3-540-85760-0_72",

language = "English (US)",

isbn = "3540857591",

pages = "577--584",

booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

address = "Germany",

}

TY - GEN

T1 - Cue Integration for medical image annotation

AU - Tommasi, Tatiana

AU - Orabona, Francesco

AU - Caputo, Barbara

N1 - Generated from Scopus record by KAUST IRTS on 2023-09-25

PY - 2008/1/1

Y1 - 2008/1/1

N2 - This paper presents the algorithms and results of our participation to the image annotation task of ImageCLEFmed 2007. We proposed a multi-cue approach where images are represented both by global and local descriptors. These cues are combined following two SVM-based strategies. The first algorithm, called Discriminative Accumulation Scheme (DAS), trains an SVM for each feature, and considers as output of each classifier the distance from the separating hyperplane. The final decision is taken on a linear combination of these distances. The second algorithm, that we call Multi Cue Kernel (MCK), uses a new Mercer kernel which can accept as input different features while keeping them separated. The DAS algorithm obtained a score of 29.9, which ranked fifth among all submissions. The MCK algorithm with the one-vs-all and with the one-vs-one multiclass extensions of SVM scored respectively 26.85 and 27.54. These runs ranked first and second among all submissions. © 2008 Springer-Verlag Berlin Heidelberg.

AB - This paper presents the algorithms and results of our participation to the image annotation task of ImageCLEFmed 2007. We proposed a multi-cue approach where images are represented both by global and local descriptors. These cues are combined following two SVM-based strategies. The first algorithm, called Discriminative Accumulation Scheme (DAS), trains an SVM for each feature, and considers as output of each classifier the distance from the separating hyperplane. The final decision is taken on a linear combination of these distances. The second algorithm, that we call Multi Cue Kernel (MCK), uses a new Mercer kernel which can accept as input different features while keeping them separated. The DAS algorithm obtained a score of 29.9, which ranked fifth among all submissions. The MCK algorithm with the one-vs-all and with the one-vs-one multiclass extensions of SVM scored respectively 26.85 and 27.54. These runs ranked first and second among all submissions. © 2008 Springer-Verlag Berlin Heidelberg.

UR - https://link.springer.com/10.1007/978-3-540-85760-0_72

UR - http://www.scopus.com/inward/record.url?scp=70349828889&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-85760-0_72

DO - 10.1007/978-3-540-85760-0_72

M3 - Conference contribution

SN - 3540857591

SP - 577

EP - 584

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -

Cue Integration for medical image annotation

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this