Ontology type: schema:Chapter Open Access: True
2019-10-10
AUTHORSSaeid Asgari Taghanaki , Aicha Bentaieb , Anmol Sharma , S. Kevin Zhou , Yefeng Zheng , Bogdan Georgescu , Puneet Sharma , Zhoubing Xu , Dorin Comaniciu , Ghassan Hamarneh
ABSTRACTSkip connections in deep networks have improved both segmentation and classification performance by facilitating the training of deeper network architectures and reducing the risks for vanishing gradients. The skip connections equip encoder-decoder like networks with richer feature representations, but at the cost of higher memory usage, computation, and possibly resulting in transferring non-discriminative feature maps. In this paper, we focus on improving the skip connections used in segmentation networks. We propose light, learnable skip connections which learn to first select the most discriminative channels, and then aggregate the selected ones as single channel attending to the most discriminative regions of input. We evaluate the proposed method on 3 different 2D and volumetric datasets and demonstrate that the proposed skip connections can outperform the traditional heavy skip connections of 4 different models in terms of segmentation accuracy (2% Dice), memory usage (at least 50%), and the number of network parameters (up to 70%). More... »
PAGES417-425
Machine Learning in Medical Imaging
ISBN
978-3-030-32691-3
978-3-030-32692-0
http://scigraph.springernature.com/pub.10.1007/978-3-030-32692-0_48
DOIhttp://dx.doi.org/10.1007/978-3-030-32692-0_48
DIMENSIONShttps://app.dimensions.ai/details/publication/pub.1121621807
JSON-LD is the canonical representation for SciGraph data.
TIP: You can open this SciGraph record using an external JSON-LD service: JSON-LD Playground Google SDTT
[
{
"@context": "https://springernature.github.io/scigraph/jsonld/sgcontext.json",
"about": [
{
"id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/08",
"inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/",
"name": "Information and Computing Sciences",
"type": "DefinedTerm"
},
{
"id": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/0801",
"inDefinedTermSet": "http://purl.org/au-research/vocabulary/anzsrc-for/2008/",
"name": "Artificial Intelligence and Image Processing",
"type": "DefinedTerm"
}
],
"author": [
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada",
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Taghanaki",
"givenName": "Saeid Asgari",
"id": "sg:person.013754250667.08",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013754250667.08"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada",
"id": "http://www.grid.ac/institutes/grid.61971.38",
"name": [
"Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada"
],
"type": "Organization"
},
"familyName": "Bentaieb",
"givenName": "Aicha",
"id": "sg:person.012771740525.65",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.012771740525.65"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada",
"id": "http://www.grid.ac/institutes/grid.61971.38",
"name": [
"Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada"
],
"type": "Organization"
},
"familyName": "Sharma",
"givenName": "Anmol",
"id": "sg:person.07537734265.49",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.07537734265.49"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Zhou",
"givenName": "S. Kevin",
"id": "sg:person.01372425362.30",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01372425362.30"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Zheng",
"givenName": "Yefeng",
"id": "sg:person.0767211426.21",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0767211426.21"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Georgescu",
"givenName": "Bogdan",
"id": "sg:person.0703547214.37",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0703547214.37"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Sharma",
"givenName": "Puneet",
"id": "sg:person.01154725330.99",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01154725330.99"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Xu",
"givenName": "Zhoubing",
"id": "sg:person.013364364655.13",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.013364364655.13"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA",
"id": "http://www.grid.ac/institutes/grid.415886.6",
"name": [
"Medical Imaging Technologies, Siemens Healthineers, Princeton, NJ, USA"
],
"type": "Organization"
},
"familyName": "Comaniciu",
"givenName": "Dorin",
"id": "sg:person.01066111014.77",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.01066111014.77"
],
"type": "Person"
},
{
"affiliation": {
"alternateName": "Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada",
"id": "http://www.grid.ac/institutes/grid.61971.38",
"name": [
"Medical Image Analysis Lab, School of Computing Science, Simon Fraser University, Burnaby, Canada"
],
"type": "Organization"
},
"familyName": "Hamarneh",
"givenName": "Ghassan",
"id": "sg:person.0642774233.16",
"sameAs": [
"https://app.dimensions.ai/discover/publication?and_facet_researcher=ur.0642774233.16"
],
"type": "Person"
}
],
"datePublished": "2019-10-10",
"datePublishedReg": "2019-10-10",
"description": "Skip connections in deep networks have improved both segmentation and classification performance by facilitating the training of deeper network architectures and reducing the risks for vanishing gradients. The skip connections equip encoder-decoder like networks with richer feature representations, but at the cost of higher memory usage, computation, and possibly resulting in transferring non-discriminative feature maps. In this paper, we focus on improving the skip connections used in segmentation networks. We propose light, learnable skip connections which learn to first select the most discriminative channels, and then aggregate the selected ones as single channel attending to the most discriminative regions of input. We evaluate the proposed method on 3 different 2D and volumetric datasets and demonstrate that the proposed skip connections can outperform the traditional heavy skip connections of 4 different models in terms of segmentation accuracy (2% Dice), memory usage (at least 50%), and the number of network parameters (up\u00a0to 70%).",
"editor": [
{
"familyName": "Suk",
"givenName": "Heung-Il",
"type": "Person"
},
{
"familyName": "Liu",
"givenName": "Mingxia",
"type": "Person"
},
{
"familyName": "Yan",
"givenName": "Pingkun",
"type": "Person"
},
{
"familyName": "Lian",
"givenName": "Chunfeng",
"type": "Person"
}
],
"genre": "chapter",
"id": "sg:pub.10.1007/978-3-030-32692-0_48",
"inLanguage": "en",
"isAccessibleForFree": true,
"isPartOf": {
"isbn": [
"978-3-030-32691-3",
"978-3-030-32692-0"
],
"name": "Machine Learning in Medical Imaging",
"type": "Book"
},
"keywords": [
"skip connections",
"memory usage",
"deep network architecture",
"high memory usage",
"rich feature representations",
"deep network",
"segmentation network",
"feature representation",
"network architecture",
"feature maps",
"segmentation accuracy",
"classification performance",
"volumetric datasets",
"discriminative regions",
"network parameters",
"discriminative channels",
"network",
"usage",
"segmentation",
"architecture",
"dataset",
"single channel",
"computation",
"different models",
"representation",
"accuracy",
"connection",
"performance",
"input",
"cost",
"channels",
"maps",
"equip",
"training",
"model",
"method",
"one",
"terms",
"number",
"parameters",
"transfer",
"gradient",
"region",
"light",
"risk",
"paper"
],
"name": "Select, Attend, and Transfer: Light, Learnable Skip Connections",
"pagination": "417-425",
"productId": [
{
"name": "dimensions_id",
"type": "PropertyValue",
"value": [
"pub.1121621807"
]
},
{
"name": "doi",
"type": "PropertyValue",
"value": [
"10.1007/978-3-030-32692-0_48"
]
}
],
"publisher": {
"name": "Springer Nature",
"type": "Organisation"
},
"sameAs": [
"https://doi.org/10.1007/978-3-030-32692-0_48",
"https://app.dimensions.ai/details/publication/pub.1121621807"
],
"sdDataset": "chapters",
"sdDatePublished": "2022-05-20T07:43",
"sdLicense": "https://scigraph.springernature.com/explorer/license/",
"sdPublisher": {
"name": "Springer Nature - SN SciGraph project",
"type": "Organization"
},
"sdSource": "s3://com-springernature-scigraph/baseset/20220519/entities/gbq_results/chapter/chapter_226.jsonl",
"type": "Chapter",
"url": "https://doi.org/10.1007/978-3-030-32692-0_48"
}
]
Download the RDF metadata as: json-ld nt turtle xml License info
JSON-LD is a popular format for linked data which is fully compatible with JSON.
curl -H 'Accept: application/ld+json' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-32692-0_48'
N-Triples is a line-based linked data format ideal for batch operations.
curl -H 'Accept: application/n-triples' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-32692-0_48'
Turtle is a human-readable linked data format.
curl -H 'Accept: text/turtle' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-32692-0_48'
RDF/XML is a standard XML format for linked data.
curl -H 'Accept: application/rdf+xml' 'https://scigraph.springernature.com/pub.10.1007/978-3-030-32692-0_48'
This table displays all metadata directly associated to this object as RDF triples.
188 TRIPLES
23 PREDICATES
71 URIs
64 LITERALS
7 BLANK NODES