Cancer is one of the most life-threatening diseases in the world, and lung cancer is the leading cause of death worldwide. If not detected at an early stage, the survival rate of lung cancer patients can be very low. To treat patients in later stages, one needs to analyze the tumour region. For accurate diagnosis of lung cancer, the first step is to detect and segment the tumor. In this chapter, an approach for segmentation of a lung tumour is presented. For pre-processing of lung CT images, simple image processing like morphological operations is used, and for tumour segmentation task, a 3D convolutional neural network (CNN) is used. The CNN architecture consists of a 3D encoder block followed by 3D decoder block just like U-Net but with deformable convolution blocks. For this study, two datasets have been used; one is the online-available NSCLC Radiomics dataset, and the other is collected from an Indian local hospital. The approach proposed in this chapter is evaluated in terms of dice coefficient. This approach is able to give significant results with a dice coefficient of 77.23%.
TopIntroduction
Cancer is a deadly disease with total cases 1,92,92,789 and 99,58,133 deaths in 2020 according to GLOBOCAN report 2020. It is a disease in which abnormal cells divide in an uncontrolled manner and can spread to nearby tissues through the process of metastasis. Lung Cancer is the most common diagnosed cancer among all other types of cancer, and is the leading cause of cancer deaths worldwide, with incidence rate 11.4% (2206771 cases) and mortality rate 18% (1796144 deaths), according to GLOBOCAN report, 2020 and India lies in South-eastern Asia where total lung cancer cases in 2020 were 1,23,309 and 1,09,520 deaths due to lung cancer (Source: GLOBOCAN 2020). The statistics is shown in Figure 1. Lung Cancer patients can be diagnosed with the help of different imaging tests like, X-rays, CT-scans, PET-scans etc. but CT scan is preferred because CT scans can give information about the specific features of the tumor, including its shape, size, location and internal density and are more accurate than chest x-rays in determining the nature of the tumor.
Figure 1. Pie charts showing (a) Incidence rate and (b) Mortality rate of different cancers across worldwide across the globe, Source: (Ferlay 2020)
The treatment of Lung Cancer and survival of patients depends on several factors like its size, type, stage and patients’ response to a specific treatment. To increase the survival rate, it is important to detect the cancer at an early stage. For early-stage detection of Lung Cancer, the nodule (size less than 30mm) detection is necessary, but it is mostly diagnosed in its later stages when treatment becomes difficult and chances of survival are less. And if it is not detected in early stages then it may spread to other parts of the body via metastasis and in that case the size of tumor (size more than 30mm) increases which needs to be detected and analyzed. This will assist the doctor in decision making for a specific treatment plan, to increase the survival rate of the patient. The tumor analysis task is divided into several sub-tasks which include tumor segmentation, detection and classification. In this chapter, a 3D Convolutional Neural Network (CNN) for the initial subtask in analysis of tumor that is tumor segmentation, is proposed. For this purpose, the authors have collected lung cancer CT image data from an Indian Hospital (Tata Memorial Hospital, Mumbai, India) and also used an online accessible dataset named NSCLC Radiomics. The CT image with nodule (size<30mm) and CT image with tumor mass (size>30mm) are shown in Figure 2.
Figure 2. CT images, left showing nodule and right showing tumor mass
The rest of the chapter is organized as, first, the background and related work regarding lung data and lung image processing, is presented, second, proposed research methodology is explained, third, results are discussed, fourth, the future directions are provided and finally the chapter is concluded in fifth section.
TopA. Lung Image Datasets
Several lung image datasets are publicly available on online platforms. These datasets include data for lung or lung lobes segmentation like LObe and Lung Analysis 2011 (LOLA 11) dataset which is a part of a challenge, data for lung nodule analysis like LIDC-IDRI (Gray 2011), LNDb (Pedrosa 2019) and data for lung tumor analysis like structseg (which is a part of a challenge), NSCLC Radiomics (Aerts, 2019) etc.