Three-Dimensional Reconstruction of a CT Image under Deep Learning Algorithm to Evaluate the Application of Percutaneous Kyphoplasty in Osteoporotic Thoracolumbar Compression Fractures.
CONTRAST MEDIA & MOLECULAR IMAGING 2022;
2022:9107021. [PMID:
35919502 PMCID:
PMC9290755 DOI:
10.1155/2022/9107021]
[Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 02/28/2022] [Revised: 03/31/2022] [Accepted: 04/02/2022] [Indexed: 11/21/2022]
Abstract
In order to investigate the therapeutic evaluation of percutaneous kyphoplasty (PKP) for
the treatment of osteoporotic thoracolumbar compression fractures by three-dimensional
(3D) reconstruction of computed tomography (CT) based on the deep learning V-Net network,
the traditional V-Net was optimized first and a new and improved V-Net was proposed. The
introduced U-Net, V-Net, and convolutional neural network (CNN) were compared in this
study. Then, 106 patients with osteoporotic thoracolumbar compression fractures were
enrolled, and 128 centrums were divided into the test group with 53 cases of PKP and the
control group with 53 cases of percutaneous vertebroplasty (PVP) according to different
surgical protocols. All patients underwent CT scan based on the improved V-Net, and data
of centrum measurement indicators, pain score, and therapeutic evaluation results of the
modified Macnab were collected. The Dice coefficient of the improved V-Net was observably
higher than that of U-Net, V-Net, and CNN, while the Hausdorff distance was lower than
that of U-Net, V-Net, and CNN (P < 0.05). The anterior
height, central height, and posterior height of the centrum were significantly higher than
those in the control group after operation (3, 5, and 7 days), while the Cobb angle
of vertebral kyphosis was significantly lower than that in the control group
(P < 0.05). The score of visual analog scale (VAS)
and analgesic use score of patients in the test group were markedly lower than those in
the control group (3, 5, and 7 days after operation),
P < 0.05. Besides, the excellent and good rate of the
test group was remarkably higher than that of the control group,
P < 0.05. Hence, the improved V-Net had better quality
of segmentation and reconstruction than the traditional deep learning network. Compared
with PVP, PKP was helpful in restoring the height of the centrum in patients with
osteoporotic thoracolumbar compression fractures and correct kyphosis, with better
analgesic effect safety.
Collapse