Page 102 - TD-3-3

Tumor Discovery Effectiveness of AI imaging for lung nodules

Early lung cancer often manifests as isolated lung efficiently identify small nodules in the early stages of lung
nodules, which can vary in characteristics and outcomes. cancer. Research has demonstrated that AI systems, by
24
A screening study in high-risk populations in Chinese learning from large volumes of cases, can automatically
communities revealed a 22.9% positive rate for lung detect lung nodules, enhance sensitivity, and reduce the
25
nodules, with a malignancy rate of approximately 6.34% workload of radiologists. These AI-assisted diagnostic
and a lung cancer detection rate of 1.5%. According to the systems have shown advantages in improving the visibility
3
26
2021 World Health Organization Classification of Thoracic of lung nodules, assessing their malignant potential, and
4,5
Tumors, carcinoma in situ and atypical adenomatous accurately detecting nodules in complex lung diseases,
hyperplasia are considered precursor glandular lesions. thus providing reliable references for diagnosing primary
A study demonstrated that early diagnosis and complete lung cancer. 27
6
resection of adenocarcinoma in situ and microinvasive In this study, we analyzed the pathological results of
adenocarcinoma resulted in 10-year disease-specific all lung nodules that met the inclusion criteria, compared
survival rates of 100%, with overall survival rates of detection outcomes between the AI-assisted group and the
95.3% and 97.8%, respectively. However, the 5-year physician-reading group, and evaluated the accuracy of the
recurrence rate after complete resection of stage I invasive AI system in predicting the malignancy probability of lung
7-9
adenocarcinoma ranges from 10% to 30%. For lung nodules. The aim of this evaluation is to assess the value of
cancer presenting as solid nodules, delayed diagnosis the AI system in the qualitative diagnosis of lung nodules.
and treatment significantly reduce 10-year survival,
irrespective of the initial size and growth rate, particularly 2. Methods
for rapidly growing nodules. Therefore, early and effective 2.1. Materials
10
identification of benign and malignant lung nodules is
crucial in clinical practice. However, determining the All patients included in this study were admitted to
nature of pulmonary nodules remains a major challenge Chongqing University Three Gorges Hospital between
and a focal point of ongoing research. June 2023 and January 2024. Each patient underwent
LDCT scans, which identified lung nodules. Following this,
Low-dose computed tomography (LDCT) has proven patients proceeded to thoracic surgery for pathological
highly effective in detecting early-stage lung cancer and has confirmation of the nodules.
replaced chest X-rays as a more sensitive screening tool.
Studies indicate that LDCT screening in high-risk groups Inclusion criteria:
significantly reduces the relative risk of lung cancer-related (i) CT images displayed lung nodules with a maximum
mortality. 11,12 Consequently, LDCT screening for high-risk diameter of ≤30 mm, measured using the lung
populations is now recommended in China. 13,14 However, window.
detecting lung nodules in computed tomography (CT) (ii) Chest scans were reconstructed with a standard
images remains challenging in routine clinical settings. algorithm, featuring a layer thickness of 0.5 – 1.25 mm.
First, lung nodules can be smaller than 3 mm, making (iii) Patients had lung nodules with confirmed pathological
them difficult for physicians to detect. Second, physicians results.
15
often focus primarily on the main clinical issue, potentially (iv) In cases with multiple lung nodules, only those with
overlooking additional findings such as nodules. 16,17 Finally, confirmed pathological results were included.
the workload of physicians has dramatically increased over Exclusion criteria:
the past 15 years, largely due to the growing number of CT (i) Patients who had undergone invasive diagnostic or
scans. The increased risk of missed diagnoses for small treatment procedures before the chest CT examination,
18
nodules—driven by radiologists’ fatigue and mental stress such as puncture biopsy or radiofrequency ablation.
during mass screening of high-risk groups—underscores (ii) Patients who had received systemic chemotherapy,
the need for more efficient nodule detection methods. targeted therapy, or chest radiotherapy for tumors
Artificial intelligence (AI) offers a promising solution to located in other parts of the body before the
assist radiologists with these tasks. In recent years, AI has examination.
emerged as a significant tool across various medical fields, (iii) Patients with a history of other malignant tumors or
driving rapid advancements in precision medicine for metastatic cancer.
cancer diagnosis and treatment. 19-23 The application of AI in (iv) Patients who exhibited atelectasis, hilar lymph node
medical imaging has also notably progressed, particularly enlargement, or pleural effusion.
in improving the detection rate of lung nodules. AI systems A total of 291 patients met the inclusion criteria.
can analyze CT images using deep learning algorithms to This group consisted of 126 males and 165 females, aged

Volume 3 Issue 3 (2024) 2 doi: 10.36922/td.4178