Hepatocellular Carcinoma Histopathology Detection Using Deep Learning and XAI & App Deployment

Local Chapter Riyadh, Saudi Arabia Chapter

Coordinated bySaudi Arabia ,

Status: Completed

Project Duration: 17 Apr 2023 - 19 Jun 2023

Open Source resources available from this project

Project background.

*. Liver cancer is one of the most commonly diagnosed cancer with the sixth highest incidence rate and the third highest mortality rate globally (2020).

*. In this project, we focus on hepatocellular carcinoma (HCC), the most prevalent subtype of primary liver cancer, which is notorious for it’s poor prognosis, with a general 5-year survival rate of 20%.

*. 905,700 people were diagnosed with and 830,200 people died from liver cancer globally in 2020.

*. Liver cancer was among the top three causes of cancer death in 46 countries.

*. The number of new cases and deaths from liver cancer could rise by >55% by 2040 globally.

*. Each year in the United States, about 25,000 men and 11,000 women get liver cancer, and about 19,000 men and 9,000 women die from the disease.

*. Between January 2004 and December 2014; 4723 liver cancer cases were registered in the Saudi Cancer Registry.

*. The highest overall age-standardised incidence rate of liver cancer among Saudi males was observed in the regions of Riyadh, Najran, and Tabuk and a recent study (Alghamdi I, 2020) found HCC to be the 4th most common cancer affecting Saudi males and the 9th most common cancer affecting females with an overall age standardized incidence rate of 5.3/100 000 population (7.5/100 000 for males and 3.1/100 000 for females).

The problem.

– The incidence of HCC is expected to increase dramatically in the next 30 years in Saudi Arabia (Abdo et al. 2012)
– Early detection of liver cancer is crucial for better prognosis and treatment outcomes 
– Liver cancer often does not cause any obvious symptoms in its early stages, which makes early detection through screening tests even more important.
– When liver cancer is diagnosed at an early stage, it is more likely to be treatable with curative intent.
– In people with early-stage liver cancers who have a liver transplant, the 5-year survival rate is in the range of 60% to 70%.
– On the other hand, liver cancer that is diagnosed at a later stage may require more aggressive treatments and can have a poor prognosis.
– Therefore, it is important to identify people who are at increased risk for liver cancer and offer them appropriate screening tests as early detection and treatment can increase the chances of survival and improve outcomes.
– High accuracy, rapid and efficient machine learning models may help to improve the rate of early detection, and by proxy, early intevention
– Explainability models will improve the understanding of the diagnosis and the model’s methods for both healthcare providers and patients.
– Any liver cancer diagnosis is one too many.

Project goals.

The Omdena Saudi Arabia Chapter aims to develop a deployed app based on Deep Learning models that will predict liver cancer (hepatocellular carcinoma or HCC) from whole slide histopathology images (WSI). The project's primary goal is to accurately classify digital pathology images as healthy or malignant, with an XAI (Explanability model) in a deployed web app.With a duration of 7-weeks, this project aims to achieve:- Data Collection and Exploratory Data Analysis - Preprocessing  - Feature Extraction - Model Development and Training - Evaluate Model - App development - App deployment - Research paper for publication - Project Presentation

Project plan.

  • Week 1

    *. Introductions
    *. Team formations (ie subtask leader assignments and members selecting their preferred subtask(s) group(s).
    *. Introduction to Omdena for all members
    *. Introduction to Omdena Saudi Arabia Chapter for all members
    *. Introduction to the project (presentation for all members)
    *. Exploring 3 datasets
    *. Exploring potential ML models
    * Github, slack, notion onboarding

  • Week 2

    *. Subtask team kick off meetings with sub task assignments, and shared milestone dates
    *. Reviewing deliverables
    *. Setting sub task set weekly meetings and deliverables
    *. Assigning members stage 1 tasks to initiate activities
    *. Selecting datasets
    *. Selecting baseline and experimental models
    *. Start EDA
    *. Start modelling
    *. Start research paper outline
    *. Start presentation outline

  • Week 3

    *. Draft of research paper, presentation, github format

  • Week 4

    * Team presentation of all subtasks for evaluation by full team leading to edits and actions to improve

  • Week 5

    Review of models and xai models and presentation and research paper, model evaluation, metrics, app deployment

  • Week 6

    app deployment, submit github, research paper and presentation

  • Week 7

    In case we need extra time

Learning outcomes.

Solving a real world problem that is growing and relevant. Team building skills, communication, time-management, research and presentation methods, data analysis and feature engineering, deep learning model development and model deployment, and feeling good about learning.

Share project on: