| Literature DB >> 35477188 |
Zhipeng Su1, Wenjie Mao1, Bin Li1, Zhizhong Zheng1, Bo Yang1, Meiyu Ren1, Tieniu Song1, Haiming Feng1, Yuqi Meng1.
Abstract
BACKGROUND: Lung cancer is the cancer with the highest mortality at home and abroad at present. The detection of lung nodules is a key step to reducing the mortality of lung cancer. Artificial intelligence-assisted diagnosis system presents as the state of the art in the area of nodule detection, differentiation between benign and malignant and diagnosis of invasive subtypes, however, a validation with clinical data is necessary for further application. Therefore, the aim of this study is to evaluate the effectiveness of artificial intelligence-assisted diagnosis system in predicting the invasive subtypes of early‑stage lung adenocarcinoma appearing as pulmonary nodules.Entities:
Keywords: Adenocarcinoma; Artificial intelligence; Invasive subtypes; Lung neoplasms; Pulmonary nodule
Mesh:
Year: 2022 PMID: 35477188 PMCID: PMC9051300 DOI: 10.3779/j.issn.1009-3419.2022.102.12
Source DB: PubMed Journal: Zhongguo Fei Ai Za Zhi ISSN: 1009-3419
患者特征
Patient characteristics
| Characteristic | AAH/AIS ( | MIA ( | IAC ( |
|
| AAH: atypical adenomatous hyperplasia; AIS: adenocarcinoma | ||||
| Age (yr) | 51.5±7.3 | 52.2±9.7 | 56.9±10.0 | 0.005 |
| Gender (Male: Female) | 6:16 | 16:15 | 74:96 | 0.204 |
| Mean CT value (HU) | -552.5±157.4 | -416.5±193.4 | -104.9±221.1 | < 0.001 |
| Diameter (cm) | 0.9±0.3 | 1.0±0.4 | 1.9±0.5 | < 0.001 |
| Volume (mm3) | 373.0 (295.0) | 457.0 (800.0) | 3, 828.0 (5, 559.0) | < 0.001 |
| Malignant probability (%) | 82.1±9.1 | 83.5±7.8 | 80.5±7.5 | < 0.001 |
结节类型与浸润亚型的关系[n(%)]
The relationship between nodule type and invasive subtypes [n(%)]
| Density | Final pathology | Total | ||
| AAH/AIS | MIA | IAC | ||
| pGGN: pure ground-glass nodule; pSN: part-solid nodule; SN: solid nodule. | ||||
| pGGN | 18 (81.8) | 11 (35.5) | 13 (7.6) | 42 |
| pSN | 4 (18.2) | 20 (64.5) | 52 (30.6) | 76 |
| SN | 0 (0.0) | 0 (0.0) | 105 (61.8) | 105 |
| Total | 22 | 31 | 170 | 223 |
影像学特征与浸润亚型的关系[n(%)]
The relationship between imaging characteristics and invasive subtypes [n(%)]
| Radiological sign | Final pathology | Density | ||||||
| AAH/AIS ( | MIA ( | IAC ( | pGGN ( | pSN ( | SN ( | |||
| Vascular penetration | 19 (86.4) | 27 (87.1) | 156 (91.8) | 0.477 | 37 (88.1) | 70 (92.1) | 95 (90.5) | 0.754 |
| Pleural retraction | 2 (9.1) | 6 (19.4) | 113 (66.5) | < 0.001 | 4 (9.5) | 33 (43.4) | 84 (80.0) | < 0.001 |
| Lobulation | 2 (9.1) | 8 (25.8) | 134 (78.8) | < 0.001 | 6 (14.3) | 48 (63.2) | 90 (85.7) | < 0.001 |
| Spiculation | 1 (4.5) | 3 (9.7) | 116 (68.2) | < 0.001 | 2 (4.8) | 40 (52.6) | 78 (74.3) | < 0.001 |
图 1人工智能系统定性诊断早期肺腺癌浸润亚型与相应病理。A、B:不典型腺瘤样增生(图 1B,HE,×200); C、D:原位腺癌(图 1D,HE,×100); E、F:微浸润性腺癌(图 1F,HE,×200); G、H:浸润性腺癌(图 1H,HE,×200)。
Artificial intelligence system for qualitative diagnosis of the invasive subtypes of early-stage lung adenocarcinoma and the corresponding pathology. A, B: atypical adenomatous hyperplasia (Fig 1B, HE, ×200); C, D: adenocarcinoma in situ (Fig 1D, HE, ×100); E, F: minimally invasive adenocarcinoma (Fig 1F, HE, ×200); G, H: invasive adenocarcinoma (Fig 1H, HE, ×200).
图 2二分类结果ROC曲线
ROC curve for two-class result. ROC: receiver operating characteristic curve.
图 3三分类结果的混淆矩阵
The confusion matrix of three-class result
图 4三分类结果的ROC曲线
ROC curve for three-class result