Mode
Text Size
Log in / Sign up

CIPHER model predicts immune checkpoint inhibitor pneumonitis risk in non-small cell lung cancer patientsLung Scans Can Now Predict Dangerous Side Effect Before Treatment

AI-generated summary of the cited source, checked by automated accuracy review. How we work

Key Takeaway
Consider prospective validation of CIPHER model for ICI-pneumonitis risk prediction in NSCLC patients.

This cohort study assessed the CIPHER model for predicting immune checkpoint inhibitor-induced pneumonitis in patients with non-small cell lung cancer. The analysis included a development cohort of 254 patients, an internal validation cohort of 93 patients, and an external validation cohort of 116 patients. Baseline CT scans served as the exposure, compared against classical radiomic models.

In the internal immunotherapy cohort, the CIPHER model distinguished patients at elevated risk from those without the event, with AUCs ranging from 0.77 to 0.85. Head-to-head benchmarking showed the CIPHER AUC was 0.83, outperforming the radiomic models. In the external validation cohort, CIPHER maintained strong performance with an AUC of 0.83 and balanced accuracy of 81.7%. The radiomic model comparison yielded a DeLong p value of 0.0318.

Confusion matrix analysis for CIPHER indicated it correctly identified 80 of 96 non-ICI-P cases and 16 of 20 ICI-P cases. While the radiomic model demonstrated high sensitivity of 85.0%, its specificity was markedly lower at 45.8%. Safety data, adverse events, and discontinuations were not reported in this study.

A key limitation is that prospective validation is required. With prospective validation, CIPHER may be incorporated into routine patient management to improve outcomes, though this remains a future possibility rather than a current standard.

  • AI spots hidden lung risks in routine CT scans
  • Helps lung cancer patients starting immunotherapy
  • Not in clinics yet — still being tested

This tool could help doctors protect patients before a serious side effect strikes.

You’re about to start a new cancer treatment. It’s powerful. It’s promising. But deep inside your lungs, a silent risk may already be brewing — one that could stop treatment in its tracks.

Doctors can’t always see it coming. Until now.

Lung cancer is one of the most common cancers worldwide. Many patients now get immunotherapy — drugs that help the immune system attack cancer. These drugs, called immune checkpoint inhibitors (ICIs), can extend lives.

But sometimes, the immune system gets too active. It starts attacking healthy lung tissue. This causes a condition called pneumonitis. In severe cases, patients must stop treatment — and some end up in the hospital.

About 3% to 5% of patients on ICIs develop this lung problem. That may sound low, but with thousands on these drugs, it adds up. And right now, doctors have no reliable way to predict who’s at risk.

They wait and watch. By the time symptoms appear — cough, shortness of breath — damage may already be done.

The Hidden Risk

What if we could see the danger before treatment even starts?

The Old Assumption

For years, doctors thought lung damage from immunotherapy was random. They believed CT scans taken before treatment looked “normal” in most patients. If the lungs appeared clear, they assumed the risk was low.

But what if the scan shows more than we can see?

Human eyes miss subtle patterns. Tiny changes in lung tissue — too faint to notice — might signal future trouble.

Here’s the twist:

A new AI tool can detect those hidden signs. It doesn’t need a special scan. Just the routine CT already taken.

Imagine your lungs are like a sponge — full of tiny air pockets. When healthy, the pattern is even and smooth. But early damage — even invisible to doctors — might create tiny irregularities.

The AI, called CIPHER, learns what these patterns look like. It’s trained on hundreds of thousands of lung images — like studying a language no human can speak.

Think of it like a smoke detector that senses heat before flames appear. CIPHER reads the CT scan and says: This lung pattern may overreact to immunotherapy.

It doesn’t diagnose disease. It predicts risk — like a weather forecast for the lungs.

This doesn’t mean this treatment is available yet.

In a study of over 300 lung cancer patients, CIPHER analyzed standard CT scans taken before immunotherapy. It correctly flagged most who later developed pneumonitis.

In testing, it was right about 83 out of 100 patients — a strong score. It found 16 of the 20 patients who got sick, and correctly cleared 80 of 96 who stayed healthy.

Another tool, based on older methods, found more sick patients — but wrongly flagged many healthy ones. That kind of false alarm can lead to unnecessary worry or missed treatment.

CIPHER was more balanced. It caught real risk without crying wolf.

This is where things get interesting.

The AI didn’t just work in one hospital. It was tested in a second group of patients — at a different center, with different machines. It performed just as well.

That’s rare in AI research. Many tools fail when moved outside their original setting. CIPHER held up.

It also beat older prediction methods — including radiomics, which pulls data from pixels. Those models were sensitive but too jumpy. CIPHER was both accurate and reliable.

This isn’t about replacing doctors. It’s about giving them better tools.

Radiologists spend minutes — sometimes hours — reviewing scans. But even experts can’t process thousands of subtle patterns at once.

AI can. And when combined with clinical judgment, it could change how we plan treatment.

Researchers say this model could one day be built into hospital systems — like a silent co-pilot.

It wouldn’t make decisions. But it could raise a flag: This patient may need extra monitoring.

If you or a loved one has lung cancer and is starting immunotherapy, this news may feel personal. You want the best shot at beating cancer — without unexpected setbacks.

Right now, CIPHER is not available in clinics. It’s still in the research phase. You can’t ask your doctor to run this test yet.

But it’s a sign of what’s coming: smarter, more personalized care. One day, your routine scan might do double duty — checking cancer and predicting risks.

For now, talk to your doctor about your personal risk for side effects. Ask about symptoms to watch for. And know that science is moving toward earlier, safer care.

The Fine Print

The study was strong — but small. Fewer than 450 patients were involved. And all had non-small cell lung cancer.

We don’t yet know if it works for other cancers or types of immunotherapy. The model also hasn’t been tested in real time — predicting risk before it happens in live patients.

Most AI tools like this never make it to hospitals. They fail in larger trials or don’t fit into clinical workflows.

So while results are promising, they’re just the first step.

What Comes Next

Researchers plan a larger, prospective study — following patients in real time. If CIPHER still performs well, it could be tested in multiple hospitals across the country.

Regulatory approval would take more time. But if all goes well, tools like this could be part of standard care in a few years.

For now, the message is clear: The future of cancer care may not come from a new drug — but from seeing more in the scans we already have.

Larger trials will test CIPHER in real-world settings, with the goal of turning routine images into early warning systems for at-risk patients.

Study Details

Study typeCohort
Sample sizen = 254
EvidenceLevel 3
PublishedApr 2026
View Original Abstract ↓
Background: Immune checkpoint inhibitors (ICIs) have revolutionized cancer therapy but can cause serious immune-related adverse events (irAEs), with pneumonitis (ICI-P) being among the most severe. Early identification of high-risk patients before ICI initiation is critical for closer monitoring, timely intervention, and improved outcomes. Purpose: To develop and validate a deep learning foundation model to predict ICI-P from baseline CT scans in patients with lung cancer. Methods: We designed the Checkpoint-Inhibitor Pneumonitis Hazard EstimatoR (CIPHER), a deep learning foundation model that combines contrastive learning with a transformer-based masked autoencoder to predict ICI-P from baseline CT scans in patients with lung cancer. Using self-supervised learning, CIPHER was pre-trained on 590,284 CT slices from 2,500 non-small cell lung cancer (NSCLC) patients to capture heterogeneous lung parenchymal patterns. After pre-training, the model was fine-tuned on an internal NSCLC cohort for ICI-P risk prediction, using images from 254 patients for model development and 93 patients for internal validation. We compared CIPHER with classical radiomic models and further evaluated it on an external NSCLC cohort of 116 patients. Results: In the internal immunotherapy cohort, CIPHER consistently distinguished patients at elevated risk of ICI-P from those without the event, with AUCs ranging from 0.77 to 0.85. In head-to-head benchmarking, CIPHER achieved an AUC of 0.83, outperforming the radiomic models. In the external validation cohort, CIPHER maintained strong performance (AUC = 0.83; balanced accuracy = 81.7%), exceeding the radiomic models (DeLong p = 0.0318) and demonstrating higher specificity without sacrificing sensitivity. By contrast, the radiomic model showed high sensitivity (85.0%) but markedly lower specificity (45.8%). Confusion matrix analysis confirmed the robust classification performance of CIPHER, correctly identifying 80 of 96 non-ICI-P cases and 16 of 20 ICI-P cases. Conclusions: We developed and externally validated CIPHER for predicting future risk of ICI-P from pre-treatment CT scans. With prospective validation, CIPHER may be incorporated into routine patient management to improve outcomes.
Free Newsletter

Clinical research that matters. Delivered to your inbox.

Join thousands of clinicians and researchers. No spam, unsubscribe anytime.