Detail
Article
Online article
FT
Medvik - BMC
  • Something wrong with this record ?

External validation of a convolutional neural network artificial intelligence tool to predict malignancy in pulmonary nodules

DR. Baldwin, J. Gustafson, L. Pickup, C. Arteta, P. Novotny, J. Declerck, T. Kadir, C. Figueiras, A. Sterba, A. Exell, V. Potesil, P. Holland, H. Spence, A. Clubley, E. O'Dowd, M. Clark, V. Ashford-Turner, ME. Callister, FV. Gleeson,

. 2020 ; 75 (4) : 306-312. [pub] 20200305

Language English Country Great Britain

Document type Journal Article, Multicenter Study, Research Support, Non-U.S. Gov't, Validation Study

Persistent link   https://www.medvik.cz/link/bmc20025125

BACKGROUND: Estimation of the risk of malignancy in pulmonary nodules detected by CT is central in clinical management. The use of artificial intelligence (AI) offers an opportunity to improve risk prediction. Here we compare the performance of an AI algorithm, the lung cancer prediction convolutional neural network (LCP-CNN), with that of the Brock University model, recommended in UK guidelines. METHODS: A dataset of incidentally detected pulmonary nodules measuring 5-15 mm was collected retrospectively from three UK hospitals for use in a validation study. Ground truth diagnosis for each nodule was based on histology (required for any cancer), resolution, stability or (for pulmonary lymph nodes only) expert opinion. There were 1397 nodules in 1187 patients, of which 234 nodules in 229 (19.3%) patients were cancer. Model discrimination and performance statistics at predefined score thresholds were compared between the Brock model and the LCP-CNN. RESULTS: The area under the curve for LCP-CNN was 89.6% (95% CI 87.6 to 91.5), compared with 86.8% (95% CI 84.3 to 89.1) for the Brock model (p≤0.005). Using the LCP-CNN, we found that 24.5% of nodules scored below the lowest cancer nodule score, compared with 10.9% using the Brock score. Using the predefined thresholds, we found that the LCP-CNN gave one false negative (0.4% of cancers), whereas the Brock model gave six (2.5%), while specificity statistics were similar between the two models. CONCLUSION: The LCP-CNN score has better discrimination and allows a larger proportion of benign nodules to be identified without missing cancers than the Brock model. This has the potential to substantially reduce the proportion of surveillance CT scans required and thus save significant resources.

References provided by Crossref.org

000      
00000naa a2200000 a 4500
001      
bmc20025125
003      
CZ-PrNML
005      
20201222153713.0
007      
ta
008      
201125s2020 xxk f 000 0|eng||
009      
AR
024    7_
$a 10.1136/thoraxjnl-2019-214104 $2 doi
035    __
$a (PubMed)32139611
040    __
$a ABA008 $b cze $d ABA008 $e AACR2
041    0_
$a eng
044    __
$a xxk
100    1_
$a Baldwin, David R $u Respiratory Medicine, Nottingham University Hospitals, City Campus, Nottingham, UK david.baldwin@nuh.nhs.uk.
245    10
$a External validation of a convolutional neural network artificial intelligence tool to predict malignancy in pulmonary nodules / $c DR. Baldwin, J. Gustafson, L. Pickup, C. Arteta, P. Novotny, J. Declerck, T. Kadir, C. Figueiras, A. Sterba, A. Exell, V. Potesil, P. Holland, H. Spence, A. Clubley, E. O'Dowd, M. Clark, V. Ashford-Turner, ME. Callister, FV. Gleeson,
520    9_
$a BACKGROUND: Estimation of the risk of malignancy in pulmonary nodules detected by CT is central in clinical management. The use of artificial intelligence (AI) offers an opportunity to improve risk prediction. Here we compare the performance of an AI algorithm, the lung cancer prediction convolutional neural network (LCP-CNN), with that of the Brock University model, recommended in UK guidelines. METHODS: A dataset of incidentally detected pulmonary nodules measuring 5-15 mm was collected retrospectively from three UK hospitals for use in a validation study. Ground truth diagnosis for each nodule was based on histology (required for any cancer), resolution, stability or (for pulmonary lymph nodes only) expert opinion. There were 1397 nodules in 1187 patients, of which 234 nodules in 229 (19.3%) patients were cancer. Model discrimination and performance statistics at predefined score thresholds were compared between the Brock model and the LCP-CNN. RESULTS: The area under the curve for LCP-CNN was 89.6% (95% CI 87.6 to 91.5), compared with 86.8% (95% CI 84.3 to 89.1) for the Brock model (p≤0.005). Using the LCP-CNN, we found that 24.5% of nodules scored below the lowest cancer nodule score, compared with 10.9% using the Brock score. Using the predefined thresholds, we found that the LCP-CNN gave one false negative (0.4% of cancers), whereas the Brock model gave six (2.5%), while specificity statistics were similar between the two models. CONCLUSION: The LCP-CNN score has better discrimination and allows a larger proportion of benign nodules to be identified without missing cancers than the Brock model. This has the potential to substantially reduce the proportion of surveillance CT scans required and thus save significant resources.
650    _2
$a dospělí $7 D000328
650    _2
$a senioři $7 D000368
650    _2
$a algoritmy $7 D000465
650    _2
$a plocha pod křivkou $7 D019540
650    12
$a umělá inteligence $7 D001185
650    _2
$a nádorová transformace buněk $x patologie $7 D002471
650    _2
$a kohortové studie $7 D015331
650    _2
$a databáze faktografické $7 D016208
650    _2
$a časná detekce nádoru $x metody $7 D055088
650    _2
$a ženské pohlaví $7 D005260
650    _2
$a lidé $7 D006801
650    _2
$a incidence $7 D015994
650    _2
$a nádory plic $x epidemiologie $x patologie $x patofyziologie $7 D008175
650    _2
$a mužské pohlaví $7 D008297
650    _2
$a lidé středního věku $7 D008875
650    _2
$a mnohočetné plicní uzly $x epidemiologie $x patologie $x patofyziologie $7 D055613
650    _2
$a invazivní růst nádoru $x patologie $7 D009361
650    _2
$a staging nádorů $7 D009367
650    12
$a neuronové sítě $7 D016571
650    _2
$a prediktivní hodnota testů $7 D011237
650    _2
$a prognóza $7 D011379
650    _2
$a ROC křivka $7 D012372
650    _2
$a retrospektivní studie $7 D012189
650    _2
$a hodnocení rizik $7 D018570
655    _2
$a časopisecké články $7 D016428
655    _2
$a multicentrická studie $7 D016448
655    _2
$a práce podpořená grantem $7 D013485
655    _2
$a validační studie $7 D023361
700    1_
$a Gustafson, Jennifer $u Radiology, Churchill Hospital, Oxford, UK.
700    1_
$a Pickup, Lyndsey $u Optellum Ltd, Oxford, UK.
700    1_
$a Arteta, Carlos $u Optellum Ltd, Oxford, UK.
700    1_
$a Novotny, Petr $u Respiratory Medicine, Glenfield General Hospital, Leicester, UK.
700    1_
$a Declerck, Jerome $u Optellum Ltd, Oxford, UK.
700    1_
$a Kadir, Timor $u Optellum Ltd, Oxford, UK.
700    1_
$a Figueiras, Catarina $u Radiology, Churchill Hospital, Oxford, UK.
700    1_
$a Sterba, Albert $u Motol Teaching Hospital V, Prague, Czech Republic.
700    1_
$a Exell, Alan $u Oxford University Hospitals NHS Foundation Trust, Oxford, Oxfordshire, UK.
700    1_
$a Potesil, Vaclav $u Optellum Ltd, Oxford, UK.
700    1_
$a Holland, Paul $u Radiology, Nottingham University Hospitals NHS Trust, Nottingham, UK.
700    1_
$a Spence, Hazel $u Radiology, Nottingham University Hospitals NHS Trust, Nottingham, UK.
700    1_
$a Clubley, Alison $u Radiology, Nottingham University Hospitals NHS Trust, Nottingham, UK.
700    1_
$a O'Dowd, Emma $u Respiratory Medicine, Nottingham University Hospitals, City Campus, Nottingham, UK.
700    1_
$a Clark, Matthew $u Radiology, Leeds Teaching Hospitals NHS Trust, Leeds, UK.
700    1_
$a Ashford-Turner, Victoria $u Respiratory Medicine, Leeds Teaching Hospitals NHS Trust, Leeds, UK.
700    1_
$a Callister, Matthew Ej $u Respiratory Medicine, Leeds Teaching Hospitals NHS Trust, Leeds, UK.
700    1_
$a Gleeson, Fergus V $u Radiology, Churchill Hospital, Oxford, UK.
773    0_
$w MED00004515 $t Thorax $x 1468-3296 $g Roč. 75, č. 4 (2020), s. 306-312
856    41
$u https://pubmed.ncbi.nlm.nih.gov/32139611 $y Pubmed
910    __
$a ABA008 $b sig $c sign $y a $z 0
990    __
$a 20201125 $b ABA008
991    __
$a 20201222153709 $b ABA008
999    __
$a ok $b bmc $g 1599270 $s 1115811
BAS    __
$a 3
BAS    __
$a PreBMC
BMC    __
$a 2020 $b 75 $c 4 $d 306-312 $e 20200305 $i 1468-3296 $m Thorax $n Thorax $x MED00004515
LZP    __
$a Pubmed-20201125

Find record

Citation metrics

Archiving options