A 'habitat' AI model outperforms standard 2D approaches in stratifying lung adenocarcinoma risk in subsolid nodules on low-dose CT scans.
Key Details
- 1The study evaluated a 'habitat' AI model, 2D model, radiomic model, and a combined model for classifying invasiveness and grade of lung adenocarcinoma presenting as subsolid nodules on LDCT.
- 2747 patients with 834 resected lung adenocarcinomas were included, split into training, internal, and external test sets.
- 3On the external test set, the macro-average AUCs were: 2D model 0.87, habitat model 0.92, radiomic model 0.92, and combined model 0.93.
- 4Habitat imaging quantifies spatial heterogeneity by segmenting nodules into subregions based on characteristics like signal intensity.
- 5Habitat and radiomic models both significantly outperformed the traditional 2D approach.
Why It Matters
Habitat AI models offer a novel, more accurate, and interpretable tool for noninvasive risk stratification of subsolid lung nodules, which could enhance early lung cancer screening workflows and reduce interobserver variability among radiologists.
Source
AuntMinnie
Related News
•Radiology Business
Experts Urge Development of Generalist Radiology AI to Cut Costs and Improve Care
Leading scientists advocate for broader, generalist radiology AI models to overcome limitations of narrow, single-task solutions.
•AuntMinnie
General LLMs Show Promise in Detecting Critical Findings in Radiology Reports
Stanford and Mayo Clinic Arizona researchers demonstrated that LLMs like GPT-4 can categorize critical findings in radiology reports using few-shot prompting.
•AuntMinnie
Experts Outline Framework and Benefits for Generalist Radiology AI
Researchers propose key features and benefits for implementing generalist radiology AI (GRAI) frameworks over narrow AI tools.