Computer Aided Lung Cancer Diagnosis
Cancer is a disease in which the cells in the human body start to grow abnormally which can cause different abnormalities in the human body and eventually death of the subject. Lung cancer is the second most common type of cancer. There are more than hundred types of cancer but the death rate of peo
2025-06-28 16:30:53 - Adil Khan
Computer Aided Lung Cancer Diagnosis
Project Area of Specialization Artificial IntelligenceProject SummaryCancer is a disease in which the cells in the human body start to grow abnormally which can cause different abnormalities in the human body and eventually death of the subject. Lung cancer is the second most common type of cancer. There are more than hundred types of cancer but the death rate of people diagnosed by lung cancer is the highest.
The American Cancer Society’s estimates for lung cancer in the United States for 2019 are about 228,150 new cases of lung cancer will be registered among them 116,440 cases will be diagnosed in men and 111,710 in women. About 142,670 people might be die by lung cancer 76,650 men and 66,020 women in the year 2019.
Diagnosis of cancer at an early stage can increase the odds of survival in the subject. There are a number of ways to diagnose lung cancer. Some of the methods of lung cancer diagnosis are medical imaging, Sputum cytology and tissue sample biopsy. Medical images such as CT scan and MRI or X-ray scans are vastly used in identifying the tumor or the cancer nodules.
Our aim is to automate the process by providing a software system that can assist the oncologists in the diagnosis of lung cancer. Such a software system can increase the effectiveness of the diagnosis of lung cancer and help in efficient prediction of cancer in the subject in a timely manner. This will not only reduce the cost of diagnosis it will also reduce the errors caused by manual inspection of medical images such as lack of experience and limitations of human capabilities in diagnosis of lung cancer.
This software system will be able to detect lung cancer, its type and stage using the CT scan of the subject. Specifically CT scan of the lungs will be fed into the system as an input .This will reduce the cost of lung cancer diagnosis, reduce the diagnosis cost and provide results in reasonable time which will increase the chance of survival of the subject.
Developing such a system using traditional software development techniques is not feasible because traditional systems need to be explicitly coded to perform a task. We need intelligent systems that can learn from data and do not need to be explicitly programmed about every possible course of action.
Such a software system will provide highly accurate results. This effective system can be beneficial in a number of ways. The training cost of the medical staff can be reduced, improved diagnosis process, reduced human errors and biasness. It can help in early diagnosis of cancer in the subject which can improve the chance of survival of the subject.
Project Objectives- The core objective of this project is to automate the process of diagnosing lung cancer from medical images. The system will take CT scan of the lungs as input and analyze the image. The system will predict if the subject have lung cancer or not. If lung cancer is present it will then analyze its type and stage. Our aim is to improve the accuracy achieved by the current computer vision systems.
- Improving the accuracy of automated lung cancer diagnosis is also one of the main objectives of the system. This will be a remarkable contribution in the ever growing field of computer vision. Improving the accuracy will result in a system that might be deployed from lab environment to the real world.
- Reducing the training time of the medical staff is also a major objective. Currently an oncologist or radiologist is responsible for manually inspecting the medical image and identifying the abnormalities in the image to diagnose lung cancer. This technique is effective but it is completely dependent on the experience and expertise of the oncologist. It takes a great deal of time to train an oncologist and radiologist and it takes even more time in perfecting the judgement of the oncologist. Our aim is to reduce this training time by providing an automated system which will enable the oncologist to diagnose lung cancer effectively and efficiently.
- Reduction of the cost of lung cancer diagnosis is another major objective. The subject have to go through a number of tests to verify that the subject have cancer .This software system will provide an effective method to diagnose lung cancer using medical images which will reduce the cost incurred by going through other tests. This will reduce the cost and time of the lung cancer diagnosis process.
- The overall reduction in time and cost will result in an effective and early diagnosis of lung cancer which will drastically improve the odds of survival of the subject. This will also result in technical advancements in Pakistan.
Developing such a system using traditional software development techniques is not feasible because traditional systems need to be explicitly coded to perform a task. Such systems will not be able to make a prediction accurately. We need intelligent systems that can learn from data and do not need to be explicitly programmed about every possible course of action. One of the applications of artificial intelligence is machine learning. Machine learning is the scientific study of algorithms and statistical models that can learn from data and do not require explicit programming to perform a task. Machine learning models possess the capabilities to rival the human capabilities and even surpass them.
Developing a system based on machine learning can provide highly accurate results. Such an effective system can be beneficial in a number of ways. The system will not be explicitly programmed which will result is less errors. The autonomous system will provide approximately instantaneous results.
This is a research and development project so the development techniques may vary as progress is made in the project. Convolutional Neural network will be the base of this project. The aim is to develop a Neural network based on convolutional neural network and multi model image analysis to improve the accuracy of the results produced by the system.
This system will be implemented using Python 3.5 in the anaconda environment. Libraries and frameworks such as OpenCV2, Tensorflow, Theano, Keras, Sci-kit Learn, Pandas, Matplotlib, Numpy, OS, Tinkter will be used to assist in the development of the machine learning model and evaluate it. Supervised Machine learning with convolutional neural network combined with the conceptual multi model image analysis will be used to develop the system. These techniques might change as progress is made in the project.
Benefits of the Project- Reduction of cost of training medical staff. It takes a great deal of time to train an oncologist and radiologist and it takes even more time in perfecting the judgement of the oncologist and radiologist. Our aim is to reduce this training time by providing an automated system which will enable the oncologist to diagnose lung cancer effectively.
- Improving the accuracy of automated lung cancer diagnosis will be a remarkable contribution in the ever growing field of computer vision. Improving the accuracy will result in a system that might be deployed from lab environment to the real world. Artificial intelligence is the future and Pakistan need such technology to keep up with the rest of the world. It is estimated that in the next 40 years the available jobs of medical staff such as radiologists will be reduced because all the systems will be automated.
- Reduction of the cost of lung cancer diagnosis. The subject have to go through a number of tests to verify that the subject have cancer .This software system will provide an effective method to diagnose lung cancer using medical images which will reduce the cost incurred by going through other tests. This will reduce the cost and time of the lung cancer diagnosis process.
- Effective and efficient diagnosis of lung cancer that can save millions of lives. This is the major benefit.
- Machine learning Model: Our aim is to develop a distinguished machine learning model that can analyze images with unparalleled accuracy. This machine learning model will be based on supervised machine learning, convolutional neural network and multi model image analysis. This model can be deployed as an independent software or it can be integrated in an existing system. Machine learning models require high computational power therefore we will require GPUs to train them and test them.
- Diagnosis Assistant: A standalone software system will be implemented to utilize the machine learning model as we currently do not have the privileges to directly integrate the models in an existing system. This standalone application will have three components Cancer detector, cancer type classifier and cancer stage classifier. Each of this component will contain a machine learning model. This system will take the CT scan of the lungs as input and analyze that image to diagnose cancer. This software system will then show the affected area of the lungs and it will display other statistical information about the image that the medical staff can use.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 36000 | |||
| Nvidia GeForce GTX 1050 ti GPU | Equipment | 1 | 30000 | 30000 |
| Overheads | Miscellaneous | 3 | 1000 | 3000 |
| Stationary, Printing Fee | Miscellaneous | 3 | 1000 | 3000 |