Idiopathic Interstitial Pneumonia
Conditions
Keywords
Hormone sensitivity, imaging features, artificial intelligence
Brief summary
Application of artificial intelligence deep learning algorithm to analyze the relationship between hormone sensitivity of idiopathic interstitial pneumonia and imaging features of high resolution CT.
Detailed description
Methods: the medical records and chest high-resolution CT images of patients with idiopathic interstitial pneumonia admitted to the respiratory department of the Third Hospital of Peking University from June 1, 2012 to December 31, 2020 were retrospectively analyzed.Application of artificial intelligence deep learning neural convolution network method to create recognition technology of different imaging features.Including ground glass, mesh, honeycomb, nodule or consolidation, the model was established. IIP patients were divided into hormone sensitive group and hormone insensitive group according to whether the use of hormone was effective or not.Logistic regression analysis was used to analyze the correlation between statistically significant parameters and hormone sensitivity.Artificial intelligence was used to establish the correlation model between imaging features and clinical data and hormone sensitivity.
Interventions
RADIATIONhigh resolution CT
Ground glass，honeycomb，reticulation, consolidation
Sponsors
Peking University Third Hospital
Study design
Observational model
CASE_CONTROL
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to 90 Years
Healthy volunteers
No
Inclusion criteria
Clinical-pathological-radiology diagnosis of idiopathic interstitial pneumonia Hormone therapy was used; The follow-up data were complete, and the effect of hormone use could be judged.
Exclusion criteria
Lung infection disease; Heart failure; Connective tissue disease; IIP Without hormone therapy ; IIP but the follow-up data were incomplete, and the effect of hormone use could not be judged.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| clinical data and imaging feature ratios in both groups | 3-6 months after medication | clinical data including ages，gender，symptoms,signs,smoking history,complications,laboratory examination，lung function. Imaging feature including ground-glass opacity, reticulation, honeycomb and consolidation. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| the relationship between imaging feature ratios and hormone sensibility | 3-6 months after medication | Logistic regression analyzing the relationship between imaging feature ratios and hormone sensibility. |
Other
| Measure | Time frame | Description |
|---|---|---|
| development of artificial intelligence algorithm model | 3-6 months after medication | The U-net method of deep learning convolutional neural network (CNN) was used to create the recognition model of different imaging features. Imaging features include ground-glass opacity, reticulation, honeycomb and consolidation. With the area ratio of imaging features of the two groups as the input and hormone efficacy as the output, the correlation model between imaging features and hormone sensitivity was established by using artificial intelligence k nearest neighbor (KNN) algorithm and support vector machine (SVM) algorithm. |
Countries
China
Outcome results
None listed