TY - JOUR

T1 - Estimation of Apnea-Hypopnea Index Using Deep Learning on 3-D Craniofacial Scans

AU - Hanif, Umaer

AU - Leary, Eileen

AU - Schneider, Logan

AU - Paulsen, Rasmus

AU - Morse, Anne Marie

AU - Blackman, Adam

AU - Schweitzer, Paula

AU - Kushida, Clete A.

AU - Liu, Stanley

AU - Jennum, Poul

AU - Sorensen, Helge

AU - Mignot, Emmanuel

N1 - Funding Information: Dr. Morse reports personal fees from Jazz Pharmaceuticals and Rhythm Pharmaceuticals unrelated to this work. She has received research or clinical trial funding from NIH/NIMH and Jazz Pharmaceuticals unrelated to this work. Funding Information: Manuscript received November 30, 2020; revised March 30, 2021 and April 26, 2021; accepted April 30, 2021. Date of publication May 7, 2021; date of current version November 5, 2021. This work was supported in part by a grant from the Klarman Family Foundation and in part by the Technical University of Denmark, and the Danish Center for Sleep Medicine. The work of Umaer Hanif at Stanford University was supported in part by Danmark-Amerika Fondet, Vera og Carl Johan Michaelsens Legat, Reinholdt W. Jorck og Hustrus Fond, Torben og Alice Frimodts Fond, Christian og Ottilia Brorsons Rejselegat, Marie og M.B. Richters Fond, Oberstløjtnant Max Nørgaard og hustru Magda Nør- Publisher Copyright: © 2013 IEEE.

PY - 2021

Y1 - 2021

N2 - Obstructive sleep apnea (OSA) is characterized by decreased breathing events that occur through the night, with severity reported as the apnea-hypopnea index (AHI), which is associated with certain craniofacial features. In this study, we used data from 1366 patients collected as part of Stanford Technology Analytics and Genomics in Sleep (STAGES) across 11 US and Canadian sleep clinics and analyzed 3D craniofacial scans with the goal of predicting AHI, as measured using gold standard nocturnal polysomnography (PSG). First, the algorithm detects pre-specified landmarks on mesh objects and aligns scans in 3D space. Subsequently, 2D images and depth maps are generated by rendering and rotating scans by 45-degree increments. Resulting images were stacked as channels and used as input to multi-view convolutional neural networks, which were trained and validated in a supervised manner to predict AHI values derived from PSGs. The proposed model achieved a mean absolute error of 11.38 events/hour, a Pearson correlation coefficient of 0.4, and accuracy for predicting OSA of 67% using 10-fold cross-validation. The model improved further by adding patient demographics and variables from questionnaires. We also show that the model performed at the level of three sleep medicine specialists, who used clinical experience to predict AHI based on 3D scan displays. Finally, we created topographic displays of the most important facial features used by the model to predict AHI, showing importance of the neck and chin area. The proposed algorithm has potential to serve as an inexpensive and efficient screening tool for individuals with suspected OSA.

AB - Obstructive sleep apnea (OSA) is characterized by decreased breathing events that occur through the night, with severity reported as the apnea-hypopnea index (AHI), which is associated with certain craniofacial features. In this study, we used data from 1366 patients collected as part of Stanford Technology Analytics and Genomics in Sleep (STAGES) across 11 US and Canadian sleep clinics and analyzed 3D craniofacial scans with the goal of predicting AHI, as measured using gold standard nocturnal polysomnography (PSG). First, the algorithm detects pre-specified landmarks on mesh objects and aligns scans in 3D space. Subsequently, 2D images and depth maps are generated by rendering and rotating scans by 45-degree increments. Resulting images were stacked as channels and used as input to multi-view convolutional neural networks, which were trained and validated in a supervised manner to predict AHI values derived from PSGs. The proposed model achieved a mean absolute error of 11.38 events/hour, a Pearson correlation coefficient of 0.4, and accuracy for predicting OSA of 67% using 10-fold cross-validation. The model improved further by adding patient demographics and variables from questionnaires. We also show that the model performed at the level of three sleep medicine specialists, who used clinical experience to predict AHI based on 3D scan displays. Finally, we created topographic displays of the most important facial features used by the model to predict AHI, showing importance of the neck and chin area. The proposed algorithm has potential to serve as an inexpensive and efficient screening tool for individuals with suspected OSA.

KW - Apnea

KW - craniofacial scans

KW - deep learning

KW - mesh

KW - multi-view

U2 - 10.1109/JBHI.2021.3078127

DO - 10.1109/JBHI.2021.3078127

M3 - Journal article

C2 - 33961569

AN - SCOPUS:85105874541

VL - 25

SP - 4185

EP - 4194

JO - IEEE Journal of Biomedical and Health Informatics

JF - IEEE Journal of Biomedical and Health Informatics

SN - 2168-2194

IS - 11

ER -