10-year mortality predicted by deep learning on chest radiographs

RadAI Slice Newsletter

Weekly Updates in Radiology AI

Good morning, there. A deep learning model estimated skeletal muscle and fat mass from over 185,000 chest radiographs, predicting 10-year mortality.

I am struck by how this model reliably forecasts clinically meaningful outcomes using only routine CXRs and deep learning. The findings suggest that opportunistic body composition analysis may become a practical, low-cost risk stratification tool, especially in older adults. It shifts how I think about the role of radiology in preventive health and long-term care planning.

Would you trust AI-derived body composition measurements for guiding geriatric care?

Here's what you need to know about Radiology AI last week:

• Deep Learning Extracts Prognostic Body Composition from Chest X-rays

• FDA Clears Aidoc’s Comprehensive Abdominal CT Triage AI

• Unified Federal AI Policy Urged for Imaging Adoption

• Prospective Multicenter Validation of Deep Learning for Mitral Valve Disease

• Plus: 1 newly released dataset, 6 FDA approved devices & 4 new papers.

🔬 Deep Learning Extracts Prognostic Body Composition from Chest X-rays

RadAI Slice: This research leverages chest x-rays to predict long-term mortality and frailty using deep learning.

The details:

• 107,568 CXR–BIA pairs for training, 77,655 for temporal validation

• Skeletal muscle mass predicted with r=0.967, MAE 1.40 kg

• Higher predicted muscle index linked to lower 10-year mortality (HR 0.65)

• Validation on 5,932 older adults and 925 geriatric clinic patients

• Performance exceeded BMI for sarcopenia and frailty associations

Key takeaway: AI-enabled CXR-based body composition estimation is accurate, prognostically meaningful, and scalable for routine risk stratification in older adults.

🛡️ FDA Clears Aidoc’s Comprehensive Abdominal CT Triage AI

Image from: Radiology Business

RadAI Slice: FDA clearance enables unified AI triage for 14 abdominal CT indications, promising efficiency gains.

The details:

• Covers 11 new and 3 prior CT triage indications

• Mean sensitivity 97%, specificity 98% in pivotal study

• Targets emergency and outpatient abdominal CT workflow

• Can flag critical findings for radiologist review

Key takeaway: Comprehensive FDA-cleared AI tools mark a step toward integrated triage and streamlined acute abdominal imaging workflows.

🏥 Unified Federal AI Policy Urged for Imaging Adoption

Image from: Radiology Business

RadAI Slice: A national policy could overcome AI deployment barriers in radiology and beyond.

The details:

• Survey of 27 major healthcare organizations

• Varied state-level rules seen as key obstacles

• Centralized rules would support broader AI trial and integration

• Radiology cited as a prime application area

Key takeaway: National regulatory harmonization may clear a path for broader, more confident clinical adoption of AI in radiology, improving care delivery.

🩺 Prospective Multicenter Validation of Deep Learning for Mitral Valve Disease

RadAI Slice: Multicenter, external validation demonstrates AI's robustness for echocardiographic mitral valve classification.

The details:

• 6,606 TTE exams from multiple nationwide sites

• External test AUROC 0.931–0.992 by etiology

• Consistency across MR severity, image quality strata

• Detects normal, rheumatic, degenerative, prolapse, functional types

Key takeaway: Externally validated echocardiography-AI for valvular disease moves imaging AI closer to scalable, generalizable clinical utility.

RECIST-CT (2026-01-08)

Modality: CT | Focus: thorax, abdomen | Task: segmentation, measurement

• Size: 58 scans from 22 patients (~1,246 lesions annotated)

• Annotations: manual 3D instance segmentations of all tumors and lymph nodes; 82 RECIST-compliant diameter measurements

• Institutions: University of Chile, Universidad de Concepción et al.

• Availability:

  public (Zenodo link)

• Highlight: First public CT dataset with full lesion segmentations (primary/metastases/nodes) plus RECIST 1.1 measurements.

🏛️ FDA Clearances

• K252970 - Aidoc's BriefCase-Triage CARE Multi-triage CT Body AI cleared for radiologist-prioritized urgent case flagging across abdominal CT.

• K253639 - GE Healthcare secures clearance for View, a new image processing system enhancing radiology image interpretation.

• K253366 - LOGIQ Fortis ultrasound, cleared for real-time blood flow assessment and tissue visualization to support broader clinical diagnostics.

• K253370 - LOGIQ Totus ultrasound platform cleared for advanced pulsed Doppler imaging of blood flow and tissue structures.

• K253784 - 3DICOM MD Cloud gets 510(k) clearance, supporting web/cloud-based radiology image interpretation workflows.

• K253009 - DS Core Detect (Dentsply Sirona) cleared; assists clinicians with AI-based radiological image analysis.

• Explore last week's 8 radiology AI FDA approvals.

📄 Fresh Papers

• doi:10.64898/2026.01.13.26343927 - Multimodal ML models can predict continuous cognitive decline with partial generalizability across independent MRI cohorts.

• doi:10.64898/2026.01.13.26343950 - Large-cohort ML dissects how body composition and brain phenotypes independently and jointly shape cognitive aging trajectories.

• doi:10.1016/j.media.2026.103954 - A new MS lesion segmentation approach demonstrates strong generalizability across MRI domains and missing contrasts.

• doi:10.64898/2026.01.13.26344036 - Analysis of a leading FCD detection AI pipeline reveals segmentation failures and imaging feature gaps as key error sources.

• Browse 139 new radiology AI studies from last week.

📰 Everything else in Radiology AI last week

• Mandated patient access to imaging reports has driven a 30% increase in radiologist workload, raising burnout concerns.

• A large review finds limited evidence that AI reduces radiologist burnout, with some data suggesting increased workload.

• Microsoft and Bristol Myers Squibb team to improve early lung cancer detection using scalable AI-based radiology tools.

• AI analysis of abdominal CT demonstrates muscle density as a key fall risk predictor in adults 45+, enabling early intervention.

• OpenEvidence hits a $12B valuation as US physician AI adoption surges, but ECRI cites chatbots as 2026's top healthcare hazard.

• Read 17 fresh radiology AI news articles for last week.

That's it for today!

👋 Quick favor: drag this into your Primary tab so you don’t miss next week. Or just hit Reply with one thought.

See you next week.

P.S. We keep building free tools to accelerate your radiology work. What's the most time-consuming pain point in your day that we should help speed up? Reply and share your take so we keep building around you.