2022 Fiscal Year Final Research Report
Development of Bone Suppression processing for thoracic radiographs of companion animals using deep learning
Project/Area Number |
20K15673
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 42020:Veterinary medical science-related
|
Research Institution | Hokkaido University |
Principal Investigator |
Shimbo Genya 北海道大学, 獣医学研究院, 特任助教 (10839252)
|
Project Period (FY) |
2020-04-01 – 2023-03-31
|
Keywords | Bone suppression / 深層学習 / 伴侶動物 / X線 |
Outline of Final Research Achievements |
Thoracic CT images from 356 cases of cats were retrieved and the signal values of bones were converted to 0. The average intensity projection images of CT scans before and after the signal value conversion were used as pseudo thoracic radiographs and pseudo bone-suppression thoracic radiographs. Then, a convolutional neural network (CNN) was trained to create a learning model that suppresses bone opacity from pseudo thoracic radiographs. Thoracic radiographs of cats were downsampled and input to the pre-trained CNN, and the output images were upsampled to the resolution of the original images. Using the aforementioned methodology, we successfully generated bone-suppressed thoracic radiographs of cats.
|
Free Research Field |
伴侶動物画像診断
|
Academic Significance and Societal Importance of the Research Achievements |
精度評価は未実施ではあるものの、本研究では動物において過去に適用困難であった胸部X線画像から骨陰影を除去するBone Suppression処理を深層学習を用いて開発した。この画像処理を用いることによりX線読影時の肺野の視認性が向上し、胸部疾患の診断精度の向上に寄与することが期待される。CT検査に全身麻酔が要求される伴侶動物臨床においては、胸部疾患の診断をX線検査に頼らざるを得ない場面に頻繁に遭遇する。したがって、本研究によって得られた画像処理は伴侶動物臨床における強力な診断支援ツールとなり得る。
|