Plusman's AI System Shows 41% Improvement in Lung Nodule Detection in Independent Clinical Study
AI-Assisted Non-Specialists Achieve Detection Rates Exceeding Specialists Without AI—Validated on Real-World Low-Dose CT Screening
TOKYO, Japan – November 11, 2025 – Plusman LLC, a Tokyo-based medical AI company, today announced significant clinical validation results demonstrating that its Plus.Lung.Nodule AI system improves lung nodule detection accuracy across all physician experience levels while maintaining specificity—validated on real-world low-dose CT screening conditions.
The comprehensive study, conducted by researchers at Nagasaki University Hospital and presented at the 2025 Japanese Lung Cancer Society Annual Conference, represents important independent validation of AI performance under the challenging imaging conditions where these systems will actually be deployed in clinical practice. The retrospective reader study analyzed 75 low-dose CT cases (61 with nodules totaling 196 nodules, 14 without) evaluated by 9 readers comprising 4 radiology specialists and 5 non-specialists. With AI assistance, per-case sensitivity improved from 87.8% to 93.8% (p<0.0001) and per-nodule sensitivity increased from 52.3% to 73.8% (p<0.0001), representing a 41% relative improvement in nodule detection capability. Notably, AI-assisted non-specialists achieved per-case sensitivity of 93.4%, exceeding specialists without AI (91.0%). Specificity remained stable at approximately 90% (p=0.51), confirming no increase in false positives.
"This study validates AI assistance using low-dose CT screening images, which represent the real-world conditions where these tools will be deployed. AI assistance significantly improved detection sensitivity across all readers regardless of their years of experience in image interpretation. Notably, specificity was maintained, meaning AI did not increase false positives or radiologist workload. As low-dose CT lung cancer screening becomes more widespread, this study provides important evidence that AI assistance has the potential to serve as a practical tool."
Key Study Highlights
Validated on Real-World Screening Conditions
The study's use of low-dose CT screening protocols—rather than standard-dose research imaging—represents validation under the challenging conditions where AI will actually be deployed in clinical practice. Low-dose CT contains significantly more noise than standard-dose imaging, making nodule detection more difficult.
Improvements Across All Experience Levels
The study demonstrated consistent improvements regardless of physician experience. Specialists improved per-nodule sensitivity from 61.4% to 78.7% (+28% improvement), while non-specialists showed even greater gains from 45.1% to 71.6% (+59% improvement). Most notably, AI-assisted non-specialists achieved performance exceeding specialists without AI assistance.
Universal Performance Across Nodule Types
Performance improvements were observed across all clinically relevant nodule classifications: solid nodules improved from 51.9% to 72.1% (+39%), pure ground-glass nodules from 44.8% to 73.5% (+64%), while part-solid nodules maintained high sensitivity at 97.4%.
Flexible Clinical Integration
The study evaluated two implementation approaches: concurrent reader (AI during initial interpretation) and second reader (AI for quality assurance). Both methods demonstrated effectiveness, confirming that implementation can be adapted to each institution's specific workflow preferences.
Conservative Study Design
The study employed a conservative methodology where readers interpreted cases with AI before reading without AI, which may inflate baseline performance due to memory effects. This suggests the published improvement values may underestimate the actual clinical benefit.
"Having AI assistance validated through independent clinical research is deeply rewarding. These results demonstrate our ability to support radiologists and contribute to early detection for patients. We remain committed to developing products that deliver real clinical value in everyday practice."
About Plus.Lung.Nodule
Plus.Lung.Nodule is a certified medical device software that analyzes CT images and assists physicians in image interpretation. The system is deployed at medical institutions throughout Japan.
- Generic Name: Workstation Program for General-Purpose Diagnostic Imaging Equipment
- Product Name: Plus.Lung.Nodule
- Japanese Certification Number: 301AGBZX00004000
About Plusman LLC
Founded in 2018 by actuaries, Plusman LLC is a Tokyo-based medical AI company advancing innovation in diagnostic imaging through artificial intelligence. The company develops a comprehensive portfolio of AI solutions including Plus.Lung.Nodule for chest CT analysis, Plus.CXR for chest X-ray analysis, and generative AI applications. Beyond the Japanese market, Plusman is actively pursuing international expansion in the United States, Europe, and Asia, contributing to enhanced healthcare quality and efficiency through AI innovation.
For more information, visit https://plusmanllc.co.jp/
Media Contact:
Hanzawa, Plusman LLC
Contact Form
Study Reference: Fukushima F, Murayama S, Tsutsui N, et al. "Impact of AI-Assisted Diagnostic Software on Readers in Low-Dose CT Lung Nodule Detection," presented at the Japanese Lung Cancer Society Annual Conference 2025.