APSR2026-915

 
Late-Breaking
Submitted
Oral Presentation
AO32-002
 
 
https://storage.unitedwebnetwork.com/files/1285/1241118-46903-APSR AO32-002_439722.pptx
Format: PPTX Only
 
Format: PDF Only
05. Respiratory Infections (non-tuberculous)
6

No more than 15 authors can be listed.
Please ensure the authors are listed in the right order.
If there are multiple affiliations, please use the "+" button next to "Select Country/Region" to add additional affiliation fields.

Ping Wu * Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
Shan You - Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
Xin Cao - Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
Wei Luo - Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
Shuxiang Yang - Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
Xiangqi Chen - Fujian Medical University Union Hospital Department of Pulmonary and Critical Care Medicine FuZhou China
 
 
 
 
 
 
 
 
 
Presenting Author Details
Ping
Wu
pingwu_19@163.com
+8615159685608
Under 30
China
Original Research
Hierarchical differentiation of clinically challenging pulmonary nodules based on CT-based deep learning radiomics model
Maximum 25 words
Abstract Content
Pulmonary adenocarcinoma, nodular cryptococcosis, and inflammatory pseudotumor may present with overlapping computed tomography (CT) manifestations, often appearing as solitary pulmonary nodules or masses with similar morphological features. Accurate differentiation remains challenging in clinical practice and may affect subsequent management strategies. This study aimed to develop and evaluate a two-step CT-based fusion model integrating deep learning, radiomics, and clinical features for differential diagnosis among these lesions.
Patients with pathologically confirmed pulmonary adenocarcinoma, nodular cryptococcosis, or inflammatory pseudotumor who underwent CT examination between January 2015 and December 2025 were retrospectively enrolled. A total of 749 patients were included, comprising 398 pulmonary adenocarcinomas, 188 nodular cryptococcoses, and 163 inflammatory pseudotumors. Patients were randomly divided into training and validation cohorts at a ratio of 7:3. Clinical characteristics, radiomics features, and two-dimensional deep learning features extracted from the largest lesion slice using ResNet101 were integrated to construct fusion models. A two-step diagnostic strategy was developed: the first model differentiated pulmonary adenocarcinoma from benign lesions, and the second model further distinguished nodular cryptococcosis from inflammatory pseudotumor.
The fusion model demonstrated robust diagnostic performance in both steps. In the first step, differentiating pulmonary adenocarcinoma from benign lesions, the model achieved an AUC of 0.92 (95% CI: 0.87–0.95) in the training cohort and 0.88 (95% CI: 0.80–0.90) in the validation cohort. In the second step, differentiating nodular cryptococcosis from inflammatory pseudotumor, the model achieved an AUC of 0.84 (95% CI: 0.80–0.89) and 0.77 (95% CI: 0.68–0.86) in the training and validation cohorts, respectively.
The proposed two-step fusion model integrating deep learning, radiomics, and clinical features provides effective differential diagnosis of pulmonary adenocarcinoma, nodular cryptococcosis, and inflammatory pseudotumor. This strategy may facilitate clinical decision-making and reduce unnecessary invasive procedures.
 
 
 
 
 
No
No
 
JPG or JPEG or PNG format ONLY
Pulmonary nodules, deep learning, hierarchical diagnosis
maximum 3 keywords and no abbreviations, separated by commas
 
297