Adrenal Adenoma AI

AI for incidental adrenal lesions in CT.

Modality:

CT

Pathology:

Adrenal lesions

Status:

Concluded


CSC Lead: Dika


This project has been withdrawn due to limited clinical capacity. This project may be revisited at a later date but it is not currently being actively worked on.


Within increasing accessibility and demand for CT imaging in radiology, the number of incidental findings are also increasing. The incidence of incidental adrenal lesions is in the region of 5%-10% (from published studies). The challenge for radiologists is giving an accurate diagnosis/characterisation of these adrenal lesions that do not clearly demonstrate macroscopic fat. Hence we often have to recall patients for dedicated adrenal CT examinations. If we are able to use AI technology to better characterise incidental adrenal lesions as either benign, malignant or indeterminate, we will reduce our recall rates of patients.

Clinical lead(s) TBC

Rationale

Incidental findings of adrenal lesions occur relatively frequently in CT. Differentiation of the nature of these lesions would help reduce patient recall for further imaging.

Patient pathway

This project investigates incidental findings so the patient pathways vary.

Training data

Retrospective CT contrast images acquired over the last 10 years with confirmed adenoma and confirmed non-adenoma. 100 scans of each are needed in first instance.

Risks

The aim of this tool is to reduce callback for further imaging in cases of incidental findings.

Goals

A CAD AI tool used on trigger by radiologist which outputs the following parameters, adenoma or non-adenoma? Then charactrise location, volume, and flag for probability of malignancy (low, medium, high). Secondary goals include information on growth on serial imaging and further characterisation of non-adenomas.

Success criteria

Accurate characterisation of adrenal lesions as adenoma or non-adenoma.

Alternatives

Currently no commercial products identified.