Description
Lynch Syndrome is associated with a significantly elevated risk of developing colorectal cancer (CRC). However, this risk can be mitigated through early detection and intervention. A pivotal diagnostic method involves identifying lesions in enterocytes using immunohistochemistry on colon tissue samples. A deficient Mismatch Repair (MMR) crypt, termed Adenomatous Crypt Foci (ACF), indicates Lynch Syndrome.
Distinguishing healthy and deficient crypts involves assessing their visual characteristics under a microscope, with healthy crypts appearing brown and ACF crypts blue. Other distinguishing criteria are explored in this project. Yet, the rarity of ACF crypts coupled with the manual microscopic examination by specialists - around 15 minutes per slide - hinders efficient diagnosis, with a single patient’s results taking around 2.5 hours due to multiple slides.
This project’s primary objective is to harness deep learning and machine learning techniques to aid specialists in detecting enterocytes and distinguishing different crypt types. This initiative aims to expedite diagnosis, facilitating swifter patient intervention during the onset of disease symptoms. Building upon the work of Clémence Lanfranchi [reference], we seek to employ advanced computational methods to enhance the diagnostic process.
The project unfolds in several phases:
- Medical Understanding: We begin by delving into the medical definition of Lynch Syndrome [reference], improving our comprehension of this condition and the intricacies of its diagnosis.
- Data Preprocessing: We explore the data preprocessing steps, detailing the crypt segmentation methodology from Clémence Lanfranchi’s work [reference]. We also outline the ranking technique employed for distinguishing crypt types. The dataset and its preprocessing modifications are introduced, addressing the challenge of limited deficient class samples.
- Experimental Protocol: We present the setup of our experimental protocol, elaborating on the classification algorithms employed and their outcomes. Deep learning methods play a crucial role in our approach.
- Future Prospects: The project concludes with a discussion on potential improvements, both in introducing advanced machine learning techniques and exploring alternative physiological-histological discriminating factors beyond the scope of Clémence Lanfranchi’s work [reference]. Additionally, we contemplate the project’s impact on environmental, societal, and ethical domains.
By integrating cutting-edge deep learning techniques into medical diagnosis, this project strives to fasten identification of Lynch Syndrome, ultimately leading to more timely and effective patient care.