Artificial Intelligence for Early Detection of Chest Nodules in X-ray Images

Abstract

Early detection increases overall survival among patients with lung cancer. This study formulated a machine learning method that processes chest X-rays (CXRs) to detect lung cancer early. After we preprocessed our dataset using monochrome and brightness correction, we used different kinds of preprocessing methods to enhance image contrast and then used U-net to perform lung segmentation. We used 559 CXRs with a single lung nodule labeled by experts to train a You Only Look Once version 4 (YOLOv4) deep-learning architecture to detect lung nodules. In a testing dataset of 100 CXRs from patients at Taipei Veterans General Hospital and 154 CXRs from the Japanese Society of Radiological Technology dataset, the sensitivity of the AI model using a combination of different preprocessing methods performed the best at 79%, with 3.04 false positives per image. We then tested the AI by using 383 sets of CXRs obtained in the past 5 years prior to lung cancer diagnoses. The median time from detection to diagnosis for radiologists assisted with AI was 46 (3–523) days, longer than that for radiologists (8 (0–263) days). The AI model can assist radiologists in the early detection of lung nodules.

Better performance of deep learning pulmonary nodule detection using chest radiography with pixel level labels in reference to computed tomography: data quality matters

“…To date, only a few studies have demonstrated the performance of deep learning algorithms in detecting and segmenting pulmonary nodule. Chiu et al 29 developed a deep learning model based on the You Only Look Once version 4 for detecting pulmonary nodules. They demonstrated that an AI model with a combination of preprocessing approaches achieved the highest sensitivity of 79%, with 3.04 false positives per image, indicating a low positive predictive value.…”

MSF-Net: A Lightweight Multi-Scale Feature Fusion Network for Skin Lesion Segmentation

“…With the development of image segmentation, more and more researchers are committed to using image segmentation in the field of medical images to assist in the diagnosis of some critical diseases including liver cancer [ 8 ], lung cancer [ 9 ], prostate cancer [ 10 ], and some kidney diseases [ 11 ]. Ronneberger et al, 2015 [ 12 ] borrowed the ideas of FCN and then proposed an end-to-end “U-shaped” network, U-Net, for medical image segmentation.…”

“…With the development of image segmentation, more and more researchers are committed to using image segmentation in the field of medical images to assist in the diagnosis of some critical diseases including liver cancer [8], lung cancer [9], prostate cancer [10], This paper proposes three modules, the S-conv Blocks, the Multi-scale Dilated Convolution Modules (MDC), and the Multi-scale Feature Fusion Modules (MFF). These modules are plug-and-play, and many ablation results show that each module improves the segmentation accuracy of the network compared to CA-Net, which is of great significance for the practical clinical application of melanoma segmentation.…”

Automatic lung segmentation in chest X-ray images using SAM with prompts from YOLO

“…Therefore, the development of CAD systems to support pulmonologists and radiologists in the assessment of chest radiographs is a priority in the respiratory community. As a core part of these systems, the task of segmenting the lung fields is critical to enable the automatic delivery of precise information about the anatomical structures identifiable in CXR images (e.g., quantification of lung nodules [18,19], detection of lung disease [20], [21] or heart failure [22]).…”