Lung Nodule Detection from Feature Engineering to Deep Learning in Thoracic CT Images: a Comprehensive Review.

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Diagnosis

Background:

• Manual lung nodule detection in CT images is time-consuming, subjective, and prone to missing small nodules.
• Variations in nodule appearance and the large number of CT slices complicate accurate identification.
• Computer-aided diagnosis (CAD) systems aim to assist radiologists by improving detection speed and accuracy.

Purpose of the Study:

• To systematically review and present the state-of-the-art in lung nodule detection using chest CT images.
• To cover published works from 2009 to April 2018, detailing various nodule detection approaches.
• To emphasize recent advancements in deep learning (DL) methods, particularly convolutional neural networks (CNNs).

Main Methods:

• Systematic literature review of research on lung nodule detection in chest CT.
• Analysis of different nodule detection methodologies.
• Focus on deep learning approaches, including various CNN architectures.

Main Results:

• Manual detection is challenging due to subjectivity, time constraints, and difficulty in identifying small or varied nodules.
• CAD systems serve as a valuable second opinion, enhancing diagnostic confidence and efficiency.
• Deep learning, especially CNNs, has become a prominent and effective approach for lung nodule detection.

Conclusions:

• The review highlights the evolution of lung nodule detection techniques over a decade.
• Deep learning-based methods, particularly CNNs, show significant promise for accurate and efficient lung nodule detection.
• Further research in DL approaches is crucial for advancing computer-aided diagnosis in radiology.