Training and validation of a deep learning U-net architecture general model for automated segmentation of inner ear from CT.
Artificial intelligence
Ear (inner)
Image processing (computer-assisted)
Neural networks (computer)
Tomography (x-ray computed)
Journal
European radiology experimental
ISSN: 2509-9280
Titre abrégé: Eur Radiol Exp
Pays: England
ID NLM: 101721752
Informations de publication
Date de publication:
12 Sep 2024
12 Sep 2024
Historique:
received:
23
02
2024
accepted:
21
08
2024
medline:
13
9
2024
pubmed:
13
9
2024
entrez:
12
9
2024
Statut:
epublish
Résumé
The intricate three-dimensional anatomy of the inner ear presents significant challenges in diagnostic procedures and critical surgical interventions. Recent advancements in deep learning (DL), particularly convolutional neural networks (CNN), have shown promise for segmenting specific structures in medical imaging. This study aimed to train and externally validate an open-source U-net DL general model for automated segmentation of the inner ear from computed tomography (CT) scans, using quantitative and qualitative assessments. In this multicenter study, we retrospectively collected a dataset of 271 CT scans to train an open-source U-net CNN model. An external set of 70 CT scans was used to evaluate the performance of the trained model. The model's efficacy was quantitatively assessed using the Dice similarity coefficient (DSC) and qualitatively assessed using a 4-level Likert score. For comparative analysis, manual segmentation served as the reference standard, with assessments made on both training and validation datasets, as well as stratified analysis of normal and pathological subgroups. The optimized model yielded a mean DSC of 0.83 and achieved a Likert score of 1 in 42% of the cases, in conjunction with a significantly reduced processing time. Nevertheless, 27% of the patients received an indeterminate Likert score of 4. Overall, the mean DSCs were notably higher in the validation dataset than in the training dataset. This study supports the external validation of an open-source U-net model for the automated segmentation of the inner ear from CT scans. This study optimized and assessed an open-source general deep learning model for automated segmentation of the inner ear using temporal CT scans, offering perspectives for application in clinical routine. The model weights, study datasets, and baseline model are worldwide accessible. A general open-source deep learning model was trained for CT automated inner ear segmentation. The Dice similarity coefficient was 0.83 and a Likert score of 1 was attributed to 42% of automated segmentations. The influence of scanning protocols on the model performances remains to be assessed.
Sections du résumé
BACKGROUND
BACKGROUND
The intricate three-dimensional anatomy of the inner ear presents significant challenges in diagnostic procedures and critical surgical interventions. Recent advancements in deep learning (DL), particularly convolutional neural networks (CNN), have shown promise for segmenting specific structures in medical imaging. This study aimed to train and externally validate an open-source U-net DL general model for automated segmentation of the inner ear from computed tomography (CT) scans, using quantitative and qualitative assessments.
METHODS
METHODS
In this multicenter study, we retrospectively collected a dataset of 271 CT scans to train an open-source U-net CNN model. An external set of 70 CT scans was used to evaluate the performance of the trained model. The model's efficacy was quantitatively assessed using the Dice similarity coefficient (DSC) and qualitatively assessed using a 4-level Likert score. For comparative analysis, manual segmentation served as the reference standard, with assessments made on both training and validation datasets, as well as stratified analysis of normal and pathological subgroups.
RESULTS
RESULTS
The optimized model yielded a mean DSC of 0.83 and achieved a Likert score of 1 in 42% of the cases, in conjunction with a significantly reduced processing time. Nevertheless, 27% of the patients received an indeterminate Likert score of 4. Overall, the mean DSCs were notably higher in the validation dataset than in the training dataset.
CONCLUSION
CONCLUSIONS
This study supports the external validation of an open-source U-net model for the automated segmentation of the inner ear from CT scans.
RELEVANCE STATEMENT
CONCLUSIONS
This study optimized and assessed an open-source general deep learning model for automated segmentation of the inner ear using temporal CT scans, offering perspectives for application in clinical routine. The model weights, study datasets, and baseline model are worldwide accessible.
KEY POINTS
CONCLUSIONS
A general open-source deep learning model was trained for CT automated inner ear segmentation. The Dice similarity coefficient was 0.83 and a Likert score of 1 was attributed to 42% of automated segmentations. The influence of scanning protocols on the model performances remains to be assessed.
Identifiants
pubmed: 39266784
doi: 10.1186/s41747-024-00508-3
pii: 10.1186/s41747-024-00508-3
doi:
Types de publication
Journal Article
Validation Study
Multicenter Study
Langues
eng
Sous-ensembles de citation
IM
Pagination
104Informations de copyright
© 2024. The Author(s).
Références
Kontorinis G, Lenarz T (2022) Superior semicircular canal dehiscence: a narrative review. J Laryngol Otol 136:284–292. https://doi.org/10.1017/S0022215121002826
doi: 10.1017/S0022215121002826
pubmed: 34615564
Dhanasingh A, Dietz A, Jolly C, Roland P (2019) Human inner-ear malformation types captured in 3D. J Int Adv Otol 15:77–82. https://doi.org/10.5152/iao.2019.6246
doi: 10.5152/iao.2019.6246
pubmed: 31058598
pmcid: 6483443
Zagzoog N, Yang VXD (2018) State of robotic mastoidectomy: literature review. World Neurosurg 116:347–351. https://doi.org/10.1016/j.wneu.2018.05.194
doi: 10.1016/j.wneu.2018.05.194
pubmed: 29870847
Auinger AB, Dahm V, Liepins R, Riss D, Baumgartner WD, Arnorldner C (2021) Robotic cochlear implant surgery: imaging-based evaluation of feasibility in clinical routine. Front Surg 8:742219. https://doi.org/10.3389/fsurg.2021.742219
doi: 10.3389/fsurg.2021.742219
pubmed: 34660683
pmcid: 8511493
Mueller F, Hermann J, Weber S, O’Toole Bom Braga G, Topsakal V (2021) Image-based planning of minimally traumatic inner ear access for robotic cochlear implantation. Front Surg 8:761217. https://doi.org/10.3389/fsurg.2021.761217
doi: 10.3389/fsurg.2021.761217
pubmed: 34901143
pmcid: 8655094
Caversaccio M, Wimmer W, Anso J et al (2019) Robotic middle ear access for cochlear implantation: First in man. PLoS One 14:e0220543. https://doi.org/10.1371/journal.pone.0220543
doi: 10.1371/journal.pone.0220543
pubmed: 31374092
pmcid: 6677292
Wang J, Liu H, Ke J et al (2020) Image-guided cochlear access by non-invasive registration: a cadaveric feasibility study. Sci Rep 10:18318. https://doi.org/10.1038/s41598-020-75530-7
doi: 10.1038/s41598-020-75530-7
pubmed: 33110188
pmcid: 7591497
Nguyen Y, Bernardeschi D, Sterkers O (2018) Potential of robot-based surgery for otosclerosis surgery. Otolaryngol Clin North Am 51:475–485. https://doi.org/10.1016/j.otc.2017.11.016
doi: 10.1016/j.otc.2017.11.016
pubmed: 29502730
Parra C, Trunet S, Granger B et al (2017) Imaging criteria to predict surgical difficulties during stapes surgery. Otol Neurotol 38:815–821. https://doi.org/10.1097/MAO.0000000000001417
doi: 10.1097/MAO.0000000000001417
pubmed: 28414695
Alenzi S, Dhanasingh A, Alanazi H, Alsanosi A, Hagr A (2021) Diagnostic value of 3D segmentation in understanding the anatomy of human inner ear including malformation types. Ear Nose Throat J 100:675S–683S. https://doi.org/10.1177/0145561320906621
doi: 10.1177/0145561320906621
pubmed: 32050777
Xianfen D, Siping C, Changhong L, Yuanmei W (2005) 3D semi-automatic segmentation of the cochlea and inner ear. Conf Proc IEEE Eng Med Biol Soc 2005:6285–6288. https://doi.org/10.1109/IEMBS.2005.1615934
doi: 10.1109/IEMBS.2005.1615934
pubmed: 17281704
Kirsch V, Nejatbakhshesfahani F, Ahmadi SA, Dieterich M, Ertl-Wagner B (2019) A probabilistic atlas of the human inner ear’s bony labyrinth enables reliable atlas-based segmentation of the total fluid space. J Neurol 266:52–61. https://doi.org/10.1007/s00415-019-09488-6
doi: 10.1007/s00415-019-09488-6
pubmed: 31422454
Vaidyanathan A, van der Lubbe MFJA, Leijenaar RTH et al (2021) Deep learning for the fully automated segmentation of the inner ear on MRI. Sci Rep 11:2885. https://doi.org/10.1038/s41598-021-82289-y
doi: 10.1038/s41598-021-82289-y
pubmed: 33536451
pmcid: 7858625
Minnema J, van Eijnatten M, Kouw W, Diblen F, Mendrik A, Wolff J (2018) CT image segmentation of bone for medical additive manufacturing using a convolutional neural network. Comput Biol Med 103:130–139. https://doi.org/10.1016/j.compbiomed.2018.10.012
doi: 10.1016/j.compbiomed.2018.10.012
pubmed: 30366309
Ding AS, Lu A, Li Z et al (2023) A self-configuring deep learning network for segmentation of temporal bone anatomy in cone-beam CT Imaging. Otolaryngol Head Neck Surg 169:988–998. https://doi.org/10.1002/ohn.317
doi: 10.1002/ohn.317
pubmed: 36883992
pmcid: 11060418
Hussain R, Lalande A, Girum KB, Guigou C, Bozorg Grayeli A (2021) Automatic segmentation of inner ear on CT-scan using auto-context convolutional neural network. Sci Rep 11:4406. https://doi.org/10.1038/s41598-021-83955-x
doi: 10.1038/s41598-021-83955-x
pubmed: 33623074
pmcid: 7902630
Zhang D, Noble JH, Dawant BM (2018) Automatic detection of the inner ears in head CT images using deep convolutional neural networks. Proc SPIE Int Soc Opt Eng 10574:1057427. https://doi.org/10.1117/12.2293383
doi: 10.1117/12.2293383
pubmed: 31007337
pmcid: 6474381
Sahiner B, Pezeshk A, Hadjiiski LM et al (2019) Deep learning in medical imaging and radiation therapy. Med Phys 46:e1–e36. https://doi.org/10.1002/mp.13264
doi: 10.1002/mp.13264
pubmed: 30367497
Yushkevich PA, Pashchinskiy A, Oguz I et al (2019) User-guided segmentation of multi-modality medical imaging datasets with ITK-SNAP. Neuroinformatics 17:83–102. https://doi.org/10.1007/s12021-018-9385-x
doi: 10.1007/s12021-018-9385-x
pubmed: 29946897
pmcid: 6310114
Perslev M, Dam EB, Pai A, Igel C (2019) One network to segment them all: a general, lightweight system for accurate 3D medical image segmentation. In: Shen D et al (eds) Medical Image Computing and Computer Assisted Intervention—MICCAI 2019. Lecture Notes in Computer Science, 11765. Springer, Cham, 30–38 https://doi.org/10.1007/978-3-030-32245-8_4
Otsu N (1979) A threshold selection method from gray-level histograms. IEEE Trans Syst Man Cybern 9:62–66. https://doi.org/10.1109/TSMC.1979.4310076
doi: 10.1109/TSMC.1979.4310076
Dice LR (1945) Measures of the amount of ecologic association between species. Ecology 26:297–302. https://doi.org/10.2307/1932409
doi: 10.2307/1932409
Mangia LRL, Coelho LOM, Carvalho B, de Oliveira AKP, Hamerschmidt R (2021) Imaging studies in otosclerosis: an up-to-date comprehensive review. Int Arch Otorhinolaryngol 25:e318–e327. https://doi.org/10.1055/s-0040-1715149
doi: 10.1055/s-0040-1715149
pubmed: 33968239
Ishiyama A, Risi F, Boyd P (2020) Potential insertion complications with cochlear implant electrodes. Cochlear Implants Int 21:206–219. https://doi.org/10.1080/14670100.2020.1730066
doi: 10.1080/14670100.2020.1730066
pubmed: 32079506
Al Saadi M, Heuninck E, De Raeve L, van de Heyning P, Topsakal V (2023) Robotic cochlear implantation in post-meningitis ossified cochlea. Am J Otolaryngol 44:103668. https://doi.org/10.1016/j.amjoto.2022.103668
doi: 10.1016/j.amjoto.2022.103668
pubmed: 36323158
Neves CA, Tran ED, Kessler IM, Blevins NH (2021) Fully automated preoperative segmentation of temporal bone structures from clinical CT scans. Sci Rep 11:116. https://doi.org/10.1038/s41598-020-80619-0
doi: 10.1038/s41598-020-80619-0
pubmed: 33420386
pmcid: 7794235
Wang J, Lv Y, Wang J et al (2021) Fully automated segmentation in temporal bone CT with neural network: a preliminary assessment study. BMC Med Imaging 21:166. https://doi.org/10.1186/s12880-021-00698-x
doi: 10.1186/s12880-021-00698-x
pubmed: 34753454
pmcid: 8576911
Ke J, Lv Y, Ma F et al (2023) Deep learning-based approach for the automatic segmentation of adult and pediatric temporal bone computed tomography images. Quant Imaging Med Surg 13:1577–1591. https://doi.org/10.21037/qims-22-658
doi: 10.21037/qims-22-658
pubmed: 36915310
pmcid: 10006112
Ding AS, Lu A, Li Z et al (2022) Automated registration-based temporal bone computed tomography segmentation for applications in neurotologic surgery. Otolaryngol Head Neck Surg 167:133–140. https://doi.org/10.1177/01945998211044982
doi: 10.1177/01945998211044982
pubmed: 34491849
Stebani J, Blaimer M, Zabler S, Neun T, Pelt DM, Rak K (2023) Towards fully automated inner ear analysis with deep learning-based joint segmentation and landmark detection framework. Sci Rep 13:19057. https://doi.org/10.1038/s41598-023-45466-9
doi: 10.1038/s41598-023-45466-9
pubmed: 37925540
pmcid: 10625555
Heutink F, Koch V, Verbist B et al (2020) Multi-scale deep learning framework for cochlea localization, segmentation and analysis on clinical ultra-high-resolution CT images. Comput Methods Programs Biomed 191:105387. https://doi.org/10.1016/j.cmpb.2020.105387
doi: 10.1016/j.cmpb.2020.105387
pubmed: 32109685
Venkatasamy A, Foll DL, Eyermann C et al (2019) Malformations of the lateral semicircular canal correlated with data from the audiogram. Eur Arch Otorhinolaryngol 276:1029–1034. https://doi.org/10.1007/s00405-019-05294-y
doi: 10.1007/s00405-019-05294-y
pubmed: 30725208
Zainol Abidin Z, Mohd Zaki F, Kew TY, Goh BS, Abdullah A (2020) Cochlear nerve canal stenosis and associated semicircular canal abnormalities in paediatric sensorineural hearing loss: a single centre study. J Laryngol Otol 134:603–609. https://doi.org/10.1017/S0022215120001334
doi: 10.1017/S0022215120001334
pubmed: 32713375
Quintas-Neves M, Saraiva J (2022) Recalling the usefulness of conebeam CT in temporal bone imaging: higher resolution with lower radiation dose. AJNR Am J Neuroradiol 43:E43–E44. https://doi.org/10.3174/ajnr.A7564
doi: 10.3174/ajnr.A7564
pubmed: 36202550
pmcid: 9731256
Benson JC, Rajendran K, Lane JI et al (2022) A new frontier in temporal bone imaging: photon-counting detector CT demonstrates superior visualization of critical anatomic structures at reduced radiation dose. AJNR Am J Neuroradiol 43:579–584. https://doi.org/10.3174/ajnr.A7452
doi: 10.3174/ajnr.A7452
pubmed: 35332019
pmcid: 8993187
Quatre R, Schmerber S, Attyé A (2024) Improving rehabilitation of deaf patients by advanced imaging before cochlear implantation. J Neurorad 51:145–154. https://doi.org/10.1016/j.neurad.2023.10.002
doi: 10.1016/j.neurad.2023.10.002
Hussain R, Guigou C, Lalande A, Bozorg Grayeli A (2022) Vision-based augmented reality system for middle ear surgery: evaluation in operating room environment. Otol Neurotol 43:385. https://doi.org/10.1097/MAO.0000000000003441
doi: 10.1097/MAO.0000000000003441
pubmed: 34889824
Braga J, Samir C, Risser L et al (2019) Cochlear shape reveals that the human organ of hearing is sex-typed from birth. Sci Rep 9:10889. https://doi.org/10.1038/s41598-019-47433-9
doi: 10.1038/s41598-019-47433-9
pubmed: 31350421
pmcid: 6659711
Bonczarowska JH, Kranioti EF (2023) Human bony labyrinth as a sex indicator in subadults. Leg Med (Tokyo) 63:102259. https://doi.org/10.1016/j.legalmed.2023.102259
doi: 10.1016/j.legalmed.2023.102259
pubmed: 37094514