NCT04121988

Diagnostic Yield of Deep Learning Based Denoising MRI in Cushing's Disease

Prospective Observational Study of Diagnostic Yield in Cushing's Disease Using Deep Learning Based Denoising MRI

• Status: Terminated
• Phase: —
• Study type: Observational
• Enrollment: 15 (actual)
• Sponsor: Asan Medical Center · Academic / Other
• Sex: All
• Age: 18 Years
• Healthy volunteers: —

Summary

Negative MRI findings may occur in up to 40% of cases of ACTH producing microadenomas. The aim of the study is to evaluate if detection of ACTH producing microadenomas can be increased using deep learning based denoising MRI.

Detailed description

Detecting ACTH producing microadenoma in MRI is important in establishing the diagnosis of Cushing disease and may enable patients to avoid additional diagnostic tests such as inferior petrosal sinus sampling. However, detecting ACTH producing microadenoma in MRI remains as a diagnostic challenge due its small size with its median diameter of 5-mm. Many attempts have been made in order to improve the sensitivity of detecting ACTH producing microadenoma. It is generally accepted as standard clinical practice to perform dynamic contrast enhanced T1 weighted image to delineate delayed enhancing microadenonoma in comparison to the background enhancement of the normal gland. Despite these attempts, negative MRI findings may occur in up to 40% of cases of ACTH producing microadenomas and there is a need to improve its detection rate. Theoretically, performing thin slice thickness scans should help detecting the lesion but this is unavoidably accompanied with increased level of noise. Deep learning based denoising algorithm can be applied to reduce the noise level and potentially increase the detection rate of ACTH producing microadenomas. The aim of the study is to evaluate if detection of ACTH producing microadenomas can be increased using deep learning based denoising MRI.

Conditions

• Pituitary ACTH Secreting Adenoma

Interventions

Type	Name	Description
DIAGNOSTIC_TEST	MRI	1 mm slice thickness with deep learning based reconstruction algorithm applied to the following sequences: * Coronal T2 weighted imaging * Dynamic contrast enhanced T1 weighted imaging * Coronal contrast enhanced T1 weighted imaging

Timeline

• Start date: 2020-01-10
• Primary completion: 2023-02-28
• Completion: 2023-02-28
• First posted: 2019-10-10
• Last updated: 2024-05-14

Locations

1 site across 1 country: South Korea