Project Description
Background
Digital pathology has been shown to improve cancer diagnosis, subtyping and grading. With the manual nature of breast cancer diagnostic routine, misdiagnosis, and inaccuracy in subtyping and grading are not uncommon. This study aims to develop an AI enabled platform that is adapted to low resource settings to aid in breast cancer diagnosis, subtyping and grading.
Methods
This will be a prospective cross-sectional study that will be carried out in the pathology laboratory of MUST and MRRH. The study will utilize 174 breast pathology specimens and specimen collection forms that will accompany the specimens to the laboratory. Data on the risk factors, presenting symptoms, and WSI will be captured annotated and uploaded in the database. These will be used to trained a machine learning model that will aid breast cancer diagnosis, subtyping and grading.
Hypothesis
At the end of the study, we expect to generate a database of breast pathology images and an AI enabled model that can diagnose, subtype and grade breast cancer
Aim
- To develop a machine learning-enabled clinicopathologic digital image platform to improve the diagnosis of breast cancer in patients presenting with breast lesions at Mbarara regional referral hospital.
Objectives
- To develop a digital platform that will be used to process and store breast lesion images, symptoms, and risk factors among women presenting at Mbarara regional referral hospital with breast lesions.
- To digitize gross and histology annotated images of breast lesions from patients presenting at Mbarara regional referral hospital
- To develop a machine learning model for diagnosis of breast cancer in patients presenting with breast lesions at Mbarara regional referral hospital