Quantitative Profiling of Lung Carcinoma Histology Images
Lead Research Organisation:
University of Warwick
Department Name: Computer Science
Abstract
Cancer contributes as a major burden of disease diagnosed around the world and is the second most common cause of death after cardiovascular disease. Due to the advancement in diagnostics and improvements in treatment and prevention of heart diseases, cancer is likely to rank as the deadliest disease in many parts of world. The overarching objective of my PhD research is to develop advanced algorithms for cancer detection, tumour segmentation, risk stratification, and survival analysis of cancer patients via analysis of whole-slide histology images.
The digitization of histology slides has recently spurred interest from computer scientists to develop techniques for the automatic detection of most commonly diagnosed cancers. These digital whole-slide images result in an explosion of data which lends itself nicely to the use of data-hungry deep learning methods to tackle digital pathology problems. During my PhD, I will mould the state-of-the-art natural image classification and segmentation methods to develop different novel approach to improve the state-of-the-art in tumour segmentation, cell classification and patient survival analyses. Those methods can generate reproducible results, reduce false positives, and carry the predictive power for cancer outcome and survival as well. Spatial analysis of different tumour and immune regions/cells at whole-slide image level will help to identify significant features that can eventually be helpful in getting better and reproducible disease outcomes. The use of deep learning methods in computational pathology could render the methods independent of datasets. Most of the methods are applicable to multiple cancer types with slight incorporation of domain information of a specific cancer. I will experiment with multiple tumour types such breast metastasis, colon, oral and lung cancers.
Outcomes of my research will be used to create a specific digital pathology image based solution that could be deployed in a pathology department with digital slide scanners. If deployed, the solution will serve as an assistant for grading of cancer and will help the pathologists in making a more reproducible and objective diagnosis and help better stratify the patients for optimizing the treatment by the oncologists.
The digitization of histology slides has recently spurred interest from computer scientists to develop techniques for the automatic detection of most commonly diagnosed cancers. These digital whole-slide images result in an explosion of data which lends itself nicely to the use of data-hungry deep learning methods to tackle digital pathology problems. During my PhD, I will mould the state-of-the-art natural image classification and segmentation methods to develop different novel approach to improve the state-of-the-art in tumour segmentation, cell classification and patient survival analyses. Those methods can generate reproducible results, reduce false positives, and carry the predictive power for cancer outcome and survival as well. Spatial analysis of different tumour and immune regions/cells at whole-slide image level will help to identify significant features that can eventually be helpful in getting better and reproducible disease outcomes. The use of deep learning methods in computational pathology could render the methods independent of datasets. Most of the methods are applicable to multiple cancer types with slight incorporation of domain information of a specific cancer. I will experiment with multiple tumour types such breast metastasis, colon, oral and lung cancers.
Outcomes of my research will be used to create a specific digital pathology image based solution that could be deployed in a pathology department with digital slide scanners. If deployed, the solution will serve as an assistant for grading of cancer and will help the pathologists in making a more reproducible and objective diagnosis and help better stratify the patients for optimizing the treatment by the oncologists.
Organisations
Publications
Shaban M
(2019)
A Novel Digital Score for Abundance of Tumour Infiltrating Lymphocytes Predicts Disease Free Survival in Oral Squamous Cell Carcinoma.
in Scientific reports
Shaban M
(2018)
Prognostic significance of automated score of tumor infiltrating lymphocytes in oral cancer.
in Journal of Clinical Oncology
Shaban M
(2020)
Context-Aware Convolutional Neural Network for Grading of Colorectal Cancer Histology Images.
in IEEE transactions on medical imaging
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/N509796/1 | 30/09/2016 | 29/09/2021 | |||
1829583 | Studentship | EP/N509796/1 | 15/11/2016 | 14/05/2020 | Muhammed Shaban |
Description | A new lymphocyte abundance based digital biomarker is developed for stratification of Oral Squamous Cell Carcinoma patients into low and high risk groups. |
Exploitation Route | Large-scale multi-centric validation can be done to establish the digital biomarker as a prognostic biomarker in Oral Squamous Cell Carcinoma. |
Sectors | Healthcare |
URL | https://tia-lab.github.io/TILAb-Score/ |