Radiology job listings still lag AI reality

Good morning, there. A new workforce analysis found that only 28% of radiology job listings mention AI or PACS technology, despite widespread deployment across imaging.

I see this as a more important signal than it may first appear. AI is already part of radiology workflow in many settings, but hiring language still treats it as background infrastructure rather than an explicit operational skill. That gap says something about where the field really is: adoption is happening, but institutional recognition is still catching up.

Has AI become part of radiology work before it became part of radiology hiring?

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

• Radiology Hiring Still Lags AI Reality

• FDA De Novo for MRI AI in Parkinsonian Syndrome

• AI Model Enables Bone Density Screening via Pediatric X-ray

• Hybrid AI Model Lifts Early Lung Cancer CT Detection

• Plus: 2 newly released datasets, 6 FDA approved devices & 4 new papers.

Would you like RadAI Slice to add a radiology AI jobs section?

📋 Radiology Hiring Still Lags AI Reality

Image from: Radiology Business

RadAISlice: A new workforce analysis highlights a disconnect between real-world imaging technology use and what radiology employers actually write into job requirements.

The details:

• RadBoard reviewed 4,000 job listings from Q1 2026

• AI or PACS was mentioned in only 28% of listings

• Just 9% specified a PACS system, and even fewer named AI tools

• The results suggest hiring language is trailing behind operational reality

Key takeaway: Radiology AI may already be routine in practice, but job descriptions still rarely reflect it. The gap suggests the field is adopting new tools faster than it is redefining the skills and systems that now shape everyday imaging work.

🧠 FDA De Novo for MRI AI in Parkinsonian Syndrome

Image from: Radiology Business

RadAISlice: The FDA granted De Novo classification to an MRI-based neuro AI tool aimed at disease differentiation in Parkinsonian syndromes.

The details:

• Designed to help distinguish Parkinson’s disease from atypical forms on MRI

• Includes conditions such as multiple system atrophy and progressive supranuclear palsy

• Uses quantitative MRI analysis to support clinical decision-making

• De Novo classification points to a novel regulatory category rather than a standard predicate-based pathway

Key takeaway: This is a notable regulatory signal for neuroimaging AI. The bigger question now is not novelty, but whether tools like this can achieve meaningful clinical adoption in difficult real-world diagnostic workflows.

🫁 AI Model Enables Bone Density Screening via Pediatric X-ray

RadAISlice: Chest x-rays combined with AI may offer a more accessible route to opportunistic bone health screening in children at risk.

The details:

• Trained on 1,464 x-rays and paired DEXA from 1,188 patients with median age 13

• External test performance reached AUC 0.90, sensitivity 82%, specificity 85%

• The model focused on spine regions associated with osteopenia risk

• Clinical outcome impact remains untested and will require further validation

Key takeaway: Opportunistic bone density assessment from routine pediatric x-rays is an appealing use case, but clinical utility will depend on prospective validation beyond retrospective performance.

🦠 Hybrid AI Model Lifts Early Lung Cancer CT Detection

Image from: EurekAlert

RadAISlice: A hybrid deep-learning architecture shows promise for more accurate CT-based early lung cancer detection.

The details:

• Reported 96%+ accuracy in distinguishing early cancer on CT

• Combines localized image features with broader contextual modeling

• Aims to improve detection while reducing review burden

• Still requires larger multi-center validation and real-world testing

Key takeaway: The technical result is promising, but this remains a familiar pattern in imaging AI: impressive performance is only the beginning, and external clinical validation is what matters next.

LSS MRI AISSLab Dataset (2026-03-26)

Modality: MRI | Focus: Lumbar spine | Task: Segmentation, Detection

• Size: 500 patients, 8,500 MRI slices, 2,979 bounding box annotations

• Annotations: Bounding boxes for foraminal stenosis with severity grades and pixel-level segmentation masks for vertebrae, discs, sacrum, posterior elements, and background.

• Institutions: Sejong University, Fırat University et al.

• Availability:

  Public (Mendeley Data)

• Highlight: A public sagittal lumbar MRI dataset combining anatomical segmentation and stenosis grading with expert clinical validation.

Dental CBCT Missing Teeth Dataset (DCBCT-MT) (2026-03-27)

Modality: CBCT | Focus: Teeth, Jawbone | Task: Detection, Classification

• Size: 158 CBCT volumes from 158 patients with 59,978 slices

• Annotations: 3D cuboid masks for missing tooth regions, slice-level JSON labels, and jaw-region classification.

• Institutions: Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, Hangzhou Dianzi University et al.

• Availability:

  Public (link)

• Highlight: A public multi-center CBCT dataset for missing tooth analysis with real-world metal artifacts and 3D annotation.

🏛️ FDA Clearances

• K252148 - Butterfly Gestational Age Tool received FDA 510(k) clearance for AI-based fetal age assessment using ultrasound.

• K252360 - Anumana's 12-lead ECG-AI algorithm received FDA clearance to support pulmonary hypertension detection.

• K253796 - Lunit INSIGHT DBT V1.2 was cleared for AI-based cancer detection in digital breast tomosynthesis.

• K253775 - SwiftMR from AIRS Medical was cleared as an automated MRI post-processing software for image analysis and segmentation.

• K253595 - Philips EPIQ and Affiniti ultrasound systems were updated with imaging improvements supporting diagnostic workflow enhancement.

• K253649 - Philips Spectral CT Veridia Family received 510(k) clearance for advanced spectral imaging and tissue characterization.

• Explore last week's 26 radiology AI FDA approvals.

📄 Fresh Papers

• doi:10.1016/j.jcct.2026.03.009 - AI-enabled cardiac volumetry on non-contrast calcium scoring CT improved prediction of atrial fibrillation in a 4,400-patient study.

• doi:10.3760/cma.j.cn112147-20251114-00713 - A Chinese expert consensus recommends combining AI-enhanced CT with pulmonary function testing for improved respiratory pathway management.

• doi:10.1007/s00247-026-06601-6 - A meta-analysis found AI-assisted ultrasound achieved 92% sensitivity and 96% specificity for infant DDH screening.

• doi:10.1186/s13244-026-02263-y - A validated open-source framework automated CNR measurement in chest CT with expert-level agreement for image quality assessment.

• Browse 157 new radiology AI studies from last week.

📰 Everything else in Radiology AI last week

• GPT-o1 outperformed Google Translate in simplifying and translating emergency radiology reports for patients.

• Capitol Imaging is using AI automation for scheduling and front-office operations in outpatient radiology.

• AuntMinnie reviewed recent AI advances and limitations in digital x-ray research, from bone health to tuberculosis detection.

• Women’s willingness to pay for AI in mammography dropped as error rates and price increased.

• A new AI model stratified which sleep apnea patients may benefit from CPAP and which may face cardiovascular risk.

• Read 11 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.