An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes

Konstantinos  Barzas; Shereen Fouad; Gainer Jasa; Gabriel Landini

An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes

Konstantinos Barzas, Shereen Fouad , Gainer Jasa, Gabriel Landini

Research output: Chapter in Book/Published conference output › Conference publication

Abstract

Cone Beam Computed Tomography (CBCT) is an indispensable imaging modality in oral radiology, offering comprehensive dental anatomical information. Accurate detection of the mandibular canal (MC), a crucial anatomical structure in the lower jaw, within CBCT volumes is essential to support clinical dentistry workflows, including diagnosis, preoperative treatment planning, and postoperative evaluation. In this study, we present a deep learning-based (DL) approach for MC segmentation using 3D U-Net and 3D Attention U-Net networks. We collected a unique dataset of CBCT scans from 20 anonymous hemisected mandibular bones, which were further processed for analysis. The samples were scanned using a CBCT scanner after inserting a wire through the whole length of the MC to identify its location in space (as a gold standard). Our experimental results demonstrate that the 3D Attention U-Net outperforms the standard 3D U-Net in detecting the MC’s location, with Dice similarity score, Precision, and Recall values of 0.65, 0.75, and 0.60, respectively. Unlike current DL-enabled methods for MC segmentation, which face deployment and trust challenges due to their ”black-box” nature, our approach incorporates a post-hoc visual explainability feature through the Grad-CAM++ (Gradient-weighted Class Activation Mapping) algorithm. This tool highlights important regions within the CBCT volumes that influence the model’s predictions, providing valuable insights into the segmentation process, and bridging the gap between cutting-edge DL technology and clinical practice.

Original language	English
Title of host publication	Computational Advances in Bio and Medical Sciences
Subtitle of host publication	12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers
Editors	Mukul S. Bansal, Wei Chen, Yury Khudyakov, Ion I. Mandoiu, Marmar R. Moussa, Murray Patterson, Sanguthevar Rajesekaran, Pavel Skums, Sharma V. Thankachan, Alexander Zelikovsky
Publisher	Springer
ISBN (Electronic)	9783031827686
ISBN (Print)	9783031827679
Publication status	Accepted/In press - 8 Nov 2023
Event	The 12th International Conference on Computational Advances in Bio and Medical Sciences - Norman, Oklahoma, United States Duration: 11 Dec 2023 → 13 Dec 2023 https://iccabs.engr.uconn.edu/

Publication series

Name	Lecture Notes in Computer Science: ICCABS proceedings
Publisher	Springer
Volume	14548
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	The 12th International Conference on Computational Advances in Bio and Medical Sciences
Abbreviated title	ICCABS 2023
Country/Territory	United States
City	Norman, Oklahoma
Period	11/12/23 → 13/12/23
Internet address	https://iccabs.engr.uconn.edu/

Bibliographical note

Keywords

Dental Cone Beam Computed Tomography · mandibular canal segmentation · U-Net deep learning model · explainable artificial intelligence · Grad-CAM

Access to Document

K Barzas et al_AAM_An Explainable Deep Learning Framework
Copyright © 2025, The Authors. This author's accepted manuscript is made available under the a Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/).
Accepted author manuscript, 786 KBLicence: CC BY 4.0

Cite this

Barzas, K., Fouad , S., Jasa, G., & Landini, G. (Accepted). An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes. In M. S. Bansal, W. Chen, Y. Khudyakov, I. I. Mandoiu, M. R. Moussa, M. Patterson, S. Rajesekaran, P. Skums, S. V. Thankachan, & A. Zelikovsky (Eds.), Computational Advances in Bio and Medical Sciences: 12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers (Lecture Notes in Computer Science: ICCABS proceedings; Vol. 14548). Springer.

Barzas, Konstantinos ; Fouad , Shereen ; Jasa, Gainer et al. / An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes. Computational Advances in Bio and Medical Sciences: 12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers. editor / Mukul S. Bansal ; Wei Chen ; Yury Khudyakov ; Ion I. Mandoiu ; Marmar R. Moussa ; Murray Patterson ; Sanguthevar Rajesekaran ; Pavel Skums ; Sharma V. Thankachan ; Alexander Zelikovsky. Springer, 2025. (Lecture Notes in Computer Science: ICCABS proceedings).

@inproceedings{4c1f51e3aef2408c92910119680a58cc,

title = "An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes",

abstract = "Cone Beam Computed Tomography (CBCT) is an indispensable imaging modality in oral radiology, offering comprehensive dental anatomical information. Accurate detection of the mandibular canal (MC), a crucial anatomical structure in the lower jaw, within CBCT volumes is essential to support clinical dentistry workflows, including diagnosis, preoperative treatment planning, and postoperative evaluation. In this study, we present a deep learning-based (DL) approach for MC segmentation using 3D U-Net and 3D Attention U-Net networks. We collected a unique dataset of CBCT scans from 20 anonymous hemisected mandibular bones, which were further processed for analysis. The samples were scanned using a CBCT scanner after inserting a wire through the whole length of the MC to identify its location in space (as a gold standard). Our experimental results demonstrate that the 3D Attention U-Net outperforms the standard 3D U-Net in detecting the MC{\textquoteright}s location, with Dice similarity score, Precision, and Recall values of 0.65, 0.75, and 0.60, respectively. Unlike current DL-enabled methods for MC segmentation, which face deployment and trust challenges due to their ”black-box” nature, our approach incorporates a post-hoc visual explainability feature through the Grad-CAM++ (Gradient-weighted Class Activation Mapping) algorithm. This tool highlights important regions within the CBCT volumes that influence the model{\textquoteright}s predictions, providing valuable insights into the segmentation process, and bridging the gap between cutting-edge DL technology and clinical practice.",

keywords = "Dental Cone Beam Computed Tomography · mandibular canal segmentation · U-Net deep learning model · explainable artificial intelligence · Grad-CAM",

author = "Konstantinos Barzas and Shereen Fouad and Gainer Jasa and Gabriel Landini",

note = "Copyright {\textcopyright} 2025, The Authors. This work is licensed under a Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/). ; The 12th International Conference on Computational Advances in Bio and Medical Sciences, ICCABS 2023 ; Conference date: 11-12-2023 Through 13-12-2023",

year = "2023",

month = nov,

day = "8",

language = "English",

isbn = "9783031827679",

series = "Lecture Notes in Computer Science: ICCABS proceedings",

publisher = "Springer",

editor = "Bansal, {Mukul S. } and Wei Chen and Yury Khudyakov and Mandoiu, {Ion I.} and Moussa, {Marmar R.} and Murray Patterson and Sanguthevar Rajesekaran and Pavel Skums and Thankachan, {Sharma V.} and Alexander Zelikovsky",

booktitle = "Computational Advances in Bio and Medical Sciences",

address = "Germany",

url = "https://iccabs.engr.uconn.edu/",

}

Barzas, K, Fouad , S, Jasa, G & Landini, G 2025, An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes. in MS Bansal, W Chen, Y Khudyakov, II Mandoiu, MR Moussa, M Patterson, S Rajesekaran, P Skums, SV Thankachan & A Zelikovsky (eds), Computational Advances in Bio and Medical Sciences: 12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers. Lecture Notes in Computer Science: ICCABS proceedings, vol. 14548, Springer, The 12th International Conference on Computational Advances in Bio and Medical Sciences, Norman, Oklahoma, United States, 11/12/23.

An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes. / Barzas, Konstantinos ; Fouad , Shereen; Jasa, Gainer et al.
Computational Advances in Bio and Medical Sciences: 12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers. ed. / Mukul S. Bansal; Wei Chen; Yury Khudyakov; Ion I. Mandoiu; Marmar R. Moussa; Murray Patterson; Sanguthevar Rajesekaran; Pavel Skums; Sharma V. Thankachan; Alexander Zelikovsky. Springer, 2025. (Lecture Notes in Computer Science: ICCABS proceedings; Vol. 14548).

Research output: Chapter in Book/Published conference output › Conference publication

TY - GEN

T1 - An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes

AU - Barzas, Konstantinos

AU - Fouad , Shereen

AU - Jasa, Gainer

AU - Landini, Gabriel

PY - 2023/11/8

Y1 - 2023/11/8

N2 - Cone Beam Computed Tomography (CBCT) is an indispensable imaging modality in oral radiology, offering comprehensive dental anatomical information. Accurate detection of the mandibular canal (MC), a crucial anatomical structure in the lower jaw, within CBCT volumes is essential to support clinical dentistry workflows, including diagnosis, preoperative treatment planning, and postoperative evaluation. In this study, we present a deep learning-based (DL) approach for MC segmentation using 3D U-Net and 3D Attention U-Net networks. We collected a unique dataset of CBCT scans from 20 anonymous hemisected mandibular bones, which were further processed for analysis. The samples were scanned using a CBCT scanner after inserting a wire through the whole length of the MC to identify its location in space (as a gold standard). Our experimental results demonstrate that the 3D Attention U-Net outperforms the standard 3D U-Net in detecting the MC’s location, with Dice similarity score, Precision, and Recall values of 0.65, 0.75, and 0.60, respectively. Unlike current DL-enabled methods for MC segmentation, which face deployment and trust challenges due to their ”black-box” nature, our approach incorporates a post-hoc visual explainability feature through the Grad-CAM++ (Gradient-weighted Class Activation Mapping) algorithm. This tool highlights important regions within the CBCT volumes that influence the model’s predictions, providing valuable insights into the segmentation process, and bridging the gap between cutting-edge DL technology and clinical practice.

AB - Cone Beam Computed Tomography (CBCT) is an indispensable imaging modality in oral radiology, offering comprehensive dental anatomical information. Accurate detection of the mandibular canal (MC), a crucial anatomical structure in the lower jaw, within CBCT volumes is essential to support clinical dentistry workflows, including diagnosis, preoperative treatment planning, and postoperative evaluation. In this study, we present a deep learning-based (DL) approach for MC segmentation using 3D U-Net and 3D Attention U-Net networks. We collected a unique dataset of CBCT scans from 20 anonymous hemisected mandibular bones, which were further processed for analysis. The samples were scanned using a CBCT scanner after inserting a wire through the whole length of the MC to identify its location in space (as a gold standard). Our experimental results demonstrate that the 3D Attention U-Net outperforms the standard 3D U-Net in detecting the MC’s location, with Dice similarity score, Precision, and Recall values of 0.65, 0.75, and 0.60, respectively. Unlike current DL-enabled methods for MC segmentation, which face deployment and trust challenges due to their ”black-box” nature, our approach incorporates a post-hoc visual explainability feature through the Grad-CAM++ (Gradient-weighted Class Activation Mapping) algorithm. This tool highlights important regions within the CBCT volumes that influence the model’s predictions, providing valuable insights into the segmentation process, and bridging the gap between cutting-edge DL technology and clinical practice.

KW - Dental Cone Beam Computed Tomography · mandibular canal segmentation · U-Net deep learning model · explainable artificial intelligence · Grad-CAM

UR - https://link.springer.com/book/9783031827679

M3 - Conference publication

SN - 9783031827679

T3 - Lecture Notes in Computer Science: ICCABS proceedings

BT - Computational Advances in Bio and Medical Sciences

A2 - Bansal, Mukul S.

A2 - Chen, Wei

A2 - Khudyakov, Yury

A2 - Mandoiu, Ion I.

A2 - Moussa, Marmar R.

A2 - Patterson, Murray

A2 - Rajesekaran, Sanguthevar

A2 - Skums, Pavel

A2 - Thankachan, Sharma V.

A2 - Zelikovsky, Alexander

PB - Springer

T2 - The 12th International Conference on Computational Advances in Bio and Medical Sciences

Y2 - 11 December 2023 through 13 December 2023

ER -

Barzas K, Fouad S, Jasa G, Landini G. An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes. In Bansal MS, Chen W, Khudyakov Y, Mandoiu II, Moussa MR, Patterson M, Rajesekaran S, Skums P, Thankachan SV, Zelikovsky A, editors, Computational Advances in Bio and Medical Sciences: 12th International Conference, ICCABS 2023, Norman, OK, USA, December 11–13, 2023, Revised Selected Papers. Springer. 2025. (Lecture Notes in Computer Science: ICCABS proceedings).

An Explainable Deep Learning Framework for Mandibular Canal Segmentation from Cone Beam Computed Tomography volumes

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this