Development of a CT-based Automated Radiomics Technique for Nodule Detection, Malignancy Prediction and Histological Subtyping of Lung Cancer.
Principal Investigator
Name
Alexander Nicol
Degrees
MPhys
Institution
The Hong Kong Polytechnic University
Position Title
Postgraduate Research Student
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-877
Initial CDAS Request Approval
Jan 26, 2022
Title
Development of a CT-based Automated Radiomics Technique for Nodule Detection, Malignancy Prediction and Histological Subtyping of Lung Cancer.
Summary
This project aims to develop an automated technique for nodule detection, malignancy prediction and cancer subtype classification which can be implemented at the screening stage. By doing so, high risk patients can be identified early, and treatment can be given promptly to improve their
prospects of survival.
Identifying the subtypes of lung cancer is critical as it affects the progression and treatment of the disease. For example, small-cell carcinoma is surgically inoperable and requires radiotherapy and chemotherapy. Other factors include the growth rate, the tendency to metastasize, and the radiosensitivity of the subtype. Therefore, prompt and accurate identification of the subtype is critical, allowing effective treatment given
early to potentially improve survival.
The novel aspects of this project are as follows: 1) The proposed technique would be, to the best of our knowledge, the first to predict the cancer status and histological subtype of lung nodules from a screening population, differentiating between benign nodules, adenocarcinoma, squamous cell carcinoma, small cell carcinoma and other cancer subtypes. Existing literature focuses on models that are trained only on patients with malignancies, usually only distinguishing between two subtypes of cancer. By developing a model that is applicable to any lung nodules identified during screening, the technique will allow for options such as patient stratification by cancer subtype, in advance of radiologist inspection of the CT images. This information could allow high risk patients to be fasttracked for earlier treatment. 2) The proposed technique would be, to the best of our knowledge, the first to combine nodule detection, segmentation, malignancy prediction and subtype classification and target a representative screening population. The knowledge gap to be addressed in this project is the use of screening CT images to identify the cancer subtypes of patients in an automated way, in order to prioritize those with the most high risk disease.
prospects of survival.
Identifying the subtypes of lung cancer is critical as it affects the progression and treatment of the disease. For example, small-cell carcinoma is surgically inoperable and requires radiotherapy and chemotherapy. Other factors include the growth rate, the tendency to metastasize, and the radiosensitivity of the subtype. Therefore, prompt and accurate identification of the subtype is critical, allowing effective treatment given
early to potentially improve survival.
The novel aspects of this project are as follows: 1) The proposed technique would be, to the best of our knowledge, the first to predict the cancer status and histological subtype of lung nodules from a screening population, differentiating between benign nodules, adenocarcinoma, squamous cell carcinoma, small cell carcinoma and other cancer subtypes. Existing literature focuses on models that are trained only on patients with malignancies, usually only distinguishing between two subtypes of cancer. By developing a model that is applicable to any lung nodules identified during screening, the technique will allow for options such as patient stratification by cancer subtype, in advance of radiologist inspection of the CT images. This information could allow high risk patients to be fasttracked for earlier treatment. 2) The proposed technique would be, to the best of our knowledge, the first to combine nodule detection, segmentation, malignancy prediction and subtype classification and target a representative screening population. The knowledge gap to be addressed in this project is the use of screening CT images to identify the cancer subtypes of patients in an automated way, in order to prioritize those with the most high risk disease.
Aims
• To develop an automated radiomics technique which uses chest CT images for
identification of malignant nodules and prediction of their histological subtype.
• To develop a patient stratification model to prioritize high risk patients for fasttracked examination and biopsy, thus facilitating early detection and prompt
treatment for better survival.
• To design a computerized tool to assist radiologists with classifying borderline
nodule cases with the aim to minimize the issues with inter-observer variability / disagreement between observers
Collaborators
Dr Shara Lee, Department of Health Technology and Informatics, The Hong Kong Polytechnic University