Antonissen et al. demonstrate that DL can reduce false positives without compromising sensitivity. Safer reassurance + earlier detection. Big step forward for #LungCancerScreening. Read more 👉

Next steps: 🧪 Prospective multicenter #trial 🌍 #Validation in incidental nodules + diverse protocols 💡 Explainable #AI tools 📋 Integration into guidelines & workflows

Not perfect: ⚠️ Still a “black box” (explainability needed) ⚠️ Retrospective, not prospective ⚠️ Mostly tested in US/EU cohorts—needs global, diverse data ⚠️ Subgroup equity (sex, race, SES) needs careful study.

Fewer false alarms → fewer unnecessary scans/biopsies → less #Scanxiety for patients → more trust in screening. Long-term, this could boost participation in #LDCT screening, now only ~18% in the US.

Clinically: At **100% sensitivity**, DL cut false positives by ~40% vs PanCan. At **100% specificity**, DL flagged ~15% of cancers for immediate referral (PanCan = none). That’s huge. #LungCancer

The DL model steps up: ✅ Higher AUROC vs PanCan ✅ Way higher AUPRC (key for rare events like cancer) ✅ Even wins when nodules are size-matched.

PanCan (aka #BrockModel) improved risk prediction using size, spiculation, and location. But size still dominates, and performance has plateaued (AUROC ~0.85–0.94). #RiskModels

Why does this matter? LDCT is an effective #preventive method (NLST, NELSON). But… ~25% of scans are “positive,” and >95% of those nodules are benign. That means anxiety, extra scans, biopsies, $$$—and lower adherence.

🚨 New in @Radiology_RSNA: A #DeepLearning model trained on NLST, tested in 3 European trials, beats the PanCan model at predicting #lung nodule #malignancy on #LDCT. Fewer false positives, same sensitivity. @radiology_rsna @RITEditor @RadiologyEditor https://t.co/MM2xe4rve7

7/ Read the full study in @Radiology_RSNA 👉 https://t.co/4lm2zzGSpR #Radiology #LungCancerScreening #CardiacImaging #OpportunisticScreening #RadInTraining @RSNA

6/ Bottom line: #Opportunistic CAC scoring on LDCT is a **powerful, low-cost, population-level tool** with the potential to transform both #LungCancer and #CVD screening. Careful steps are still necessary to standardize use, establish thresholds, and integrate it into practice.

5/ But… ⚠️ - Current prevention guidelines use CAC from gated CT. - Thresholds for action on LDCT in a population at risk for #lungCancer are unknown. ➡️ Risks: overtreatment, undertreatment, inconsistent advice - #radiologist workload: How to use reliable #AI

4/ Why it matters: - #LungCancer screening is **underutilized** in US (only %16 of 16 million eligible). - Adding heart risk info might increase enrollment💡 - One LDCT = two screenings in a population susceptible to #CVDs + low cost + effective

3/ Results? - Any CAC >0 → higher all-cause & #CVD mortality. - Dose-response effect: more CAC or more vessels involved, higher risk. - Robust even with **old, thick-slice (>5mm) CTs**. ➡️ Meaning: this works on real-world and old scans, not just pristine research protocols.

2/ 📊 Using >12,000 participants from the ELCAP cohort, CAC was scored **opportunistically** (ordinal 0–3 scale) on **baseline LDCT**. Follow-up? Nearly **23 years**. That’s *long-term* with a capital L.

Can a #LungCancer #screening #CT also predict long-term heart health and #mortality? 🫁🫀 Shemesh et al. demonstrate that a simple coronary artery calcium (CAC) score on #LDCT predicts mortality up to 25 YEARS. @radiology_rsna @radiology @RITEditor @VChernyakMD @RadiologyEditor

Humbled that @NajvaApp has hit 1000+ users! I've dictated 100K+ words with it - it's become an inseparable part of my workflow. Voice-to-text + LLM is the ultimate productivity tool. Try it yourself (always free) and feel the magic! 🎙️✨ https://t.co/R8zPtzxcjx

🧵Tweetorial: 1/ Can #AI outperform radiologists in detecting contralateral breast cancer after mastectomy? New study in @radiology_rsna puts standalone AI to the test on unilateral mammograms in women with prior mastectomy. #BreastImaging @RITEditor @VChernyakMD @DrLindaMoy

This study explores how CT imaging can help predict regional lymph node metastases in patients with MSI-high colon cancer☢️🧬🦀 Read more: https://t.co/Lcrp2WHshf @RITEditor @VChernyakMD @DrLindaMoy @DHochhegger #Radiology #CancerResearch #ColorectalCancer #CT #MSI

1/8 Ever looked for a perfect teaching case or cases for a research project? New in @radiology_rsna: RadSearch, a semantic search model trained on >16K radiology reports, designed to retrieve relevant reports more accurately than standard methods https://t.co/ewjgxeBhAM