Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Apr 3;4(4):100314.
doi: 10.1016/j.bpsgos.2024.100314. eCollection 2024 Jul.

Segmentation and Volume Estimation of the Habenula Using Deep Learning in Patients With Depression

Yusuke Kyuragi  1 , Naoya Oishi  1 , Momoko Hatakoshi  1 , Jinichi Hirano  2 , Takamasa Noda  3 , Yujiro Yoshihara  1 , Yuri Ito  1 , Hiroyuki Igarashi  1 , Jun Miyata  1   4 , Kento Takahashi  1 , Kei Kamiya  2 , Junya Matsumoto  5 , Tomohisa Okada  6 , Yasutaka Fushimi  7 , Kazuyuki Nakagome  8 , Masaru Mimura  2 , Toshiya Murai  1 , Taro Suwa  1
Affiliations

Segmentation and Volume Estimation of the Habenula Using Deep Learning in Patients With Depression

Yusuke Kyuragi et al. Biol Psychiatry Glob Open Sci. .

Abstract

Background: The habenula is involved in the pathophysiology of depression. However, its small structure limits the accuracy of segmentation methods, and the findings regarding its volume have been inconsistent. This study aimed to create a highly accurate habenula segmentation model using deep learning, test its generalizability to clinical magnetic resonance imaging, and examine differences between healthy participants and patients with depression.

Methods: This multicenter study included 382 participants (patients with depression: N = 234, women 47.0%; healthy participants: N = 148, women 37.8%). A 3-dimensional residual U-Net was used to create a habenula segmentation model on 3T magnetic resonance images. The reproducibility and generalizability of the predictive model were tested on various validation cohorts. Thereafter, differences between the habenula volume of healthy participants and that of patients with depression were examined.

Results: A Dice coefficient of 86.6% was achieved in the derivation cohort. The test-retest dataset showed a mean absolute percentage error of 6.66, indicating sufficiently high reproducibility. A Dice coefficient of >80% was achieved for datasets with different imaging conditions, such as magnetic field strengths, spatial resolutions, and imaging sequences, by adjusting the threshold. A significant negative correlation with age was observed in the general population, and this correlation was more pronounced in patients with depression (p < 10-7, r = -0.59). Habenula volume decreased with depression severity in women even when the effects of age and scanner were excluded (p = .019, η2 = 0.099).

Conclusions: Habenula volume could be a pathophysiologically relevant factor and diagnostic and therapeutic marker for depression, particularly in women.

Keywords: Deep learning; Depression; Habenula; Image analysis; Sex differences; Structural MRI.

Plain language summary

Accurate segmentation of the habenula, a brain region implicated in depression, is challenging. In this study, we developed an automated human habenula segmentation model using deep learning techniques. The model was confirmed to be reproducible and generalizable at various spatial resolutions. Application of this model to a multicenter dataset confirmed that habenula volume decreased with age in healthy volunteers, an association that was more pronounced in individuals with depression. In addition, habenula volume decreased with the severity of depression in women. This novel model for habenula segmentation enables further study of the role of the habenula in depression.

PubMed Disclaimer

Figures

Figure 2
Figure 2
Habenula segmentation using deep learning. (A) Three-dimensional (3D) residual U-Net model architecture, (B) training and testing Dice coefficient curves of the model, (C) example of prediction in 1 participant. T1WI, T1-weighted image.
Figure 1
Figure 1
Overview of data acquisition and analysis. Structural images of the brain were obtained using five 3T magnetic resonance imaging scanners at 3 facilities. Using magnetic resonance images from healthy control participants (HC) acquired using scanner 1 as derivation cohort, a prediction model was built using deep learning, and the habenula was predicted on other images from a validation cohort. A habenula volume analysis was conducted, and the generalizability of the model was verified using the predicted habenula volume. 3D Res U-Net, 3-dimensional residual U-Net; CV, cross-validation; Dep, patient with depression; MPRAGE, magnetization-prepared rapid acquisition gradient-echo; SANLM, spatial-adaptive nonlocal means; T1WI, T1-weighted image.
Figure 3
Figure 3
Volume assessment with validation dataset. (A) The test-retest dataset. The figures on the left show the correlation with the habenula volume calculated from the first and second scans in the same healthy participants. The figures on the right show the Bland-Altman plots. The predicted habenula volumes were used in the upper figure, and those corrected by total intracranial volume (TIV) in the lower figure. (B) Traveling subject dataset. The figures on the left show the correlation with the habenula volume calculated from the images acquired using the 3T and 7T scanners in the same healthy participants. The figures on the right show the Bland-Altman plots. The predicted habenula volumes were used in the upper figure, and those corrected by TIV in the lower figure. MAPE, mean absolute percentage error.
Figure 4
Figure 4
Habenula volume in the healthy control group. (A) Habenula volume according to the side (left), sex (middle), and left-right and sex (right). (B) Habenula volume divided by the total intracranial volume (TIV) according to the side (left), sex (middle), and left-right and sex (right). n.s., not significant.
Figure 5
Figure 5
Habenula volume change in patients with depression (Dep) compared with healthy control participants (HC). Habenula volume (A) according to group and sex, (B) correlation with age in both groups, (C) correlation according to group and sex, and (D) difference according to the severity of depression (left). Association of habenula volume with the 17-item Hamilton Depression Rating Scale (HDRS) scores (D, right), and (E) the difference in habenula volume according to the severity of depression by sex. Considering the brain size effect, total intracranial volume (TIV) was covaried in the model in which analysis of covariance was conducted. n.s., not significant.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Namboodiri V.M., Rodriguez-Romaguera J., Stuber G.D. The habenula. Curr Biol. 2016;26:R873–R877. - PubMed
    1. Hu H., Cui Y., Yang Y. Circuits and functions of the lateral habenula in health and in disease. Nat Rev Neurosci. 2020;21:277–295. - PubMed
    1. Li B., Piriz J., Mirrione M., Chung C., Proulx C.D., Schulz D., et al. Synaptic potentiation onto habenula neurons in the learned helplessness model of depression. Nature. 2011;470:535–539. - PMC - PubMed
    1. Li K., Zhou T., Liao L., Yang Z., Wong C., Henn F., et al. βCaMKII in lateral habenula mediates core symptoms of depression. Science. 2013;341:1016–1020. - PMC - PubMed
    1. Yang Y., Cui Y., Sang K., Dong Y., Ni Z., Ma S., Hu H. Ketamine blocks bursting in the lateral habenula to rapidly relieve depression. Nature. 2018;554:317–322. - PubMed

LinkOut - more resources