Combining Handcrafted Features and Deep Learning for Automatic Classification of Lung Cancer on CT Scans

Pallavi Deshpande; Mohammed Wasim Bhatt; Santaji Krishna Shinde; Neelam Labhade-Kumar; N. Ashokkumar; K. G. S. Venkatesan; Finney Daniel Shadrach

doi:10.37965/jait.2023.0388

Combining Handcrafted Features and Deep Learning for Automatic Classification of Lung Cancer on CT Scans

Authors

Pallavi Deshpande Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune, India https://orcid.org/0000-0002-2203-5867
Mohammed Wasim Bhatt Model Institute of Engineering and Technology Jammu, Jammu & Kashmir, India https://orcid.org/0000-0003-0542-2790
Santaji Krishna Shinde Vidya Pratishthan’s Kamalnayan Bajaj Institute of Engineering & Technology, Baramati, Pune, Maharashtra, India https://orcid.org/0000-0002-0563-2686
Neelam Labhade-Kumar Shree Ramchandra College of Engineering, Wagholi, Pune, India https://orcid.org/0000-0002-4571-1135
N. Ashokkumar Mohan Babu University, Erstwhile Sree Vidyanikethan Engineering College, Tirupati-517102, Andra Pradesh, India https://orcid.org/0000-0002-4034-484X
K. G. S. Venkatesan MEGHA Institute of Engineering & Technology for Women, Edulabad, Hyderabad, Telangana, India
Finney Daniel Shadrach KPR Institute of Engineering and Technology, Coimbatore, India https://orcid.org/0000-0003-1037-0614

DOI:

https://doi.org/10.37965/jait.2023.0388

Keywords:

CT image classification, deep learning, handcrafted features, lung cancer, lung nodule classification

Abstract

On a global scale, lung cancer is responsible for around 27% of all cancer fatalities. Even though there have been great strides in diagnosis and therapy in recent years, the five-year cure rate is just 19%. Classification is crucial for diagnosing lung nodules. This is especially true today that automated categorization may provide a professional opinion that can be used by doctors. New computer vision and machine learning techniques have made possible accurate and quick categorization of CT images. This field of research has exploded in popularity in recent years because of its high efficiency and ability to decrease labour requirements. Here, they want to look carefully at the current state of automated categorization of lung nodules. General-purpose structures are briefly discussed, and typical algorithms are described. Our results show deep learning-based lung nodule categorization quickly becomes the industry standard. Therefore, it is critical to pay greater attention to the coherence of the data inside the study and the consistency of the research topic. Furthermore, there should be greater collaboration between designers, medical experts, and others in the field.