Air Bronchogram Explained: The Chest X-ray Sign Every Radiologist Must Recognize

Introduction

Every day, thousands of chest radiographs are interpreted in emergency departments and intensive care units around the world. Although artificial intelligence has become increasingly capable of detecting pulmonary abnormalities, there remains one classic radiologic sign that every radiologist, pulmonologist, emergency physician, and critical care specialist must immediately recognize—the air bronchogram.

This seemingly simple imaging finding often represents the difference between early recognition of life-threatening alveolar disease and delayed diagnosis. Whether encountered in severe bacterial pneumonia, pulmonary edema, acute respiratory distress syndrome (ARDS), pulmonary hemorrhage, or certain lung malignancies, an air bronchogram provides direct evidence that the surrounding alveoli have lost their normal air content while the bronchi remain patent.

The clinical importance of this sign is well illustrated by the uploaded case: a 26-year-old woman with end-stage renal disease secondary to lupus nephritis who presented with fever, cough, and dyspnea. Her condition rapidly progressed to severe ARDS requiring extracorporeal membrane oxygenation (ECMO). Portable chest radiography demonstrated bilateral air-space consolidation with classic air bronchograms, and bronchoalveolar lavage cultures identified Moraxella catarrhalis and Haemophilus influenzae as the causative pathogens. Following antimicrobial therapy, renal replacement therapy, and intensive supportive care, she recovered successfully.

In this article, we examine not only the radiologic significance of air bronchograms but also their role in modern AI-assisted imaging workflows, enterprise radiology platforms, and clinical decision support systems. This combination of classic imaging interpretation and emerging artificial intelligence is transforming how critical pulmonary diseases are detected and managed.

Clinical Background

Air bronchogram is one of the oldest yet most valuable radiographic signs in thoracic imaging. Despite advances in multidetector CT, photon-counting CT, functional MRI, and deep learning–based image analysis, the recognition of an air bronchogram on a simple chest radiograph remains a cornerstone of diagnostic radiology.

An air bronchogram is visualized when air-filled bronchi become conspicuous because the surrounding alveoli are opacified by fluid, inflammatory exudate, blood, proteinaceous material, or tumor cells. Under normal physiological conditions, both bronchi and alveoli contain air, producing little contrast between them. Once the alveolar air is replaced by pathological material while the bronchi remain patent, the branching bronchial tree becomes sharply outlined against the dense pulmonary parenchyma.

This radiographic sign therefore provides more than an imaging clue—it offers insight into the underlying pathophysiology of lung disease. Recognizing whether the bronchi remain open, whether consolidation is segmental or diffuse, and whether associated findings such as pleural effusion, atelectasis, or volume loss are present allows the radiologist to narrow the differential diagnosis rapidly.

The uploaded case perfectly illustrates this principle. A young immunocompromised woman with systemic lupus erythematosus complicated by end-stage renal disease developed severe bacterial pneumonia, which rapidly progressed to acute respiratory distress syndrome (ARDS). Portable chest radiography demonstrated diffuse bilateral pulmonary consolidation with classic air bronchograms, reflecting extensive alveolar filling while the bronchial lumen remained air-filled and patent. Bronchoalveolar lavage subsequently identified Moraxella catarrhalis and Haemophilus influenzae, confirming infectious pneumonia as the underlying cause.

Patient Story: From Mild Pneumonia to Life-Threatening ARDS

The clinical course of this patient highlights how rapidly pulmonary infection can evolve in immunocompromised individuals.

A 26-year-old woman with lupus nephritis–associated end-stage renal disease presented with a three-day history of fever, productive cough, and progressive shortness of breath. Initially, her symptoms resembled community-acquired pneumonia. However, over the following hours she developed worsening hypoxemia despite supplemental oxygen.

Mechanical ventilation was initiated, but severe refractory respiratory failure ensued. Because conventional ventilatory support failed to maintain adequate oxygenation, veno-venous ECMO was established through a dual-lumen catheter placed in the right internal jugular vein, enabling extracorporeal oxygenation while minimizing ventilator-induced lung injury.

Portable chest radiography performed during this critical phase revealed diffuse bilateral air-space opacification with prominent air bronchograms—an imaging pattern strongly suggestive of widespread alveolar consolidation. Bronchoalveolar lavage cultures later grew Moraxella catarrhalis and Haemophilus influenzae, organisms well recognized as respiratory pathogens in immunocompromised hosts.

Following broad-spectrum antimicrobial therapy, renal replacement therapy, and aggressive intensive care support, the patient's pulmonary function gradually recovered. ECMO was successfully discontinued after nine days, and mechanical ventilation was weaned four days later. At six-month follow-up she remained clinically stable without significant respiratory sequelae.

This case underscores several key clinical lessons:

Immunocompromised patients may deteriorate rapidly despite initially nonspecific respiratory symptoms.
Portable chest radiography remains indispensable in ICU monitoring.
Air bronchogram is often among the earliest and most reliable radiographic indicators of alveolar consolidation.
Early imaging recognition can accelerate diagnosis, microbiologic investigation, and timely escalation to advanced therapies such as ECMO.

Pathophysiology of the Air Bronchogram

To appreciate the diagnostic value of the air bronchogram, it is essential to understand the microscopic anatomy of the lung.

The lungs consist of millions of alveoli connected to terminal bronchioles and bronchi. In healthy lungs:

Alveoli are filled with air.
Bronchi are filled with air.
Both structures therefore have nearly identical radiographic density.

Because there is little intrinsic contrast, the bronchi are generally invisible on standard chest radiographs.

Disease changes this relationship dramatically.

When the alveoli become filled with inflammatory exudate, edema fluid, hemorrhage, proteinaceous material, or tumor cells, the surrounding lung parenchyma becomes radiopaque. If the bronchi remain open and air-filled, they appear as branching, dark tubular structures coursing through dense white consolidation—the classic air bronchogram.

Common Pathologic Processes Producing Air Bronchograms

Bacterial pneumonia
Viral pneumonia
Acute respiratory distress syndrome (ARDS)
Cardiogenic pulmonary edema
Diffuse alveolar hemorrhage
Pulmonary contusion
Organizing pneumonia
Bronchioloalveolar-pattern adenocarcinoma
Pulmonary infarction
Certain inflammatory lung diseases

The presence of an air bronchogram therefore suggests alveolar filling with preserved bronchial patency, distinguishing these disorders from obstructive atelectasis, where mucus plugging or bronchial obstruction may prevent visualization of air-filled bronchi.

Imaging Findings

Figure 1. Portable Chest Radiograph Demonstrating Bilateral Air Bronchograms

Interpretation

Examination: Portable AP chest radiograph

Findings:

Extensive bilateral patchy to confluent air-space opacities
Dense alveolar consolidation involving both lungs
Multiple branching radiolucent tubular structures consistent with air bronchograms
No evidence of focal cavitation
Findings compatible with diffuse alveolar disease
ECMO catheter visualized in expected position
Imaging appearance highly suggestive of severe infectious pneumonia complicated by ARDS

Impression:

Diffuse bilateral pulmonary consolidation with extensive air bronchograms consistent with severe alveolar filling disease. Clinical correlation strongly favors acute respiratory distress syndrome secondary to bacterial pneumonia.

CT Correlation: Why Computed Tomography Matters

Although chest radiography remains the frontline imaging modality in critically ill patients, computed tomography (CT) provides substantially greater anatomic detail and plays an essential role when the diagnosis remains uncertain or complications are suspected.

CT can confirm that the branching lucencies observed on chest radiography truly represent air-filled bronchi traversing consolidated lung parenchyma, rather than vascular structures or overlapping anatomical shadows.

Compared with plain radiography, CT allows radiologists to evaluate:

Distribution of consolidation
Bronchial patency
Airway obstruction
Cavitation
Pleural disease
Mediastinal lymphadenopathy
Pulmonary embolism
Underlying malignancy

This distinction is especially important because an identical chest radiograph may represent dramatically different diseases requiring entirely different treatments.

Figure 2. Representative Chest CT Demonstrating Air Bronchograms (Different Patient).
Axial contrast-enhanced CT image demonstrates branching air-filled bronchi within dense lobar consolidation. This representative CT image is from a different patient and is included for educational purposes only to illustrate the characteristic CT appearance of an air bronchogram. The featured case in this article is based on chest radiography.

Because the featured patient underwent portable chest radiography rather than diagnostic chest CT, a representative CT image from a different patient is included to demonstrate the classic tomographic appearance of an air bronchogram and to facilitate comparison between radiographic and CT findings.

CT Appearance of an Air Bronchogram

Typical CT findings include:

✓ Air-filled bronchi surrounded by soft-tissue attenuation

✓ Segmental or lobar consolidation

✓ Preservation of bronchial caliber

✓ Absence of proximal airway obstruction

✓ Sharp contrast between bronchi and consolidated alveoli

In bacterial pneumonia, CT frequently demonstrates homogeneous consolidation with numerous branching air bronchograms.

In ARDS, the pattern is usually bilateral, diffuse, gravity dependent, and accompanied by widespread ground-glass opacity.

Differential Diagnosis of Air Bronchogram

One of the greatest strengths of an experienced thoracic radiologist is the ability to integrate the air bronchogram sign into a meaningful differential diagnosis.

The radiographic appearance should never be interpreted in isolation.

Clinical history, laboratory findings, patient age, immune status, and disease progression are equally important.

1. Bacterial Pneumonia

This is the most common cause of an air bronchogram.

Inflammatory exudate fills the alveoli while the bronchi remain patent.

Typical imaging findings include:

Segmental consolidation
Lobar consolidation
Air bronchograms
Possible parapneumonic effusion

The current case belongs to this category.

Bronchoalveolar lavage identified Moraxella catarrhalis and Haemophilus influenzae, confirming bacterial pneumonia as the underlying etiology.

2. Acute Respiratory Distress Syndrome (ARDS)

ARDS represents diffuse inflammatory injury to the alveolar-capillary membrane.

Typical imaging findings include:

Bilateral diffuse consolidation
Dependent lung opacities
Air bronchograms
Rapid progression

Unlike cardiogenic pulmonary edema, cardiomegaly is often absent.

This patient's severe respiratory failure requiring ECMO is characteristic of advanced ARDS.

3. Pulmonary Edema

Both cardiogenic and non-cardiogenic edema may demonstrate air bronchograms.

Additional findings include:

Kerley B lines
Pleural effusion
Cardiomegaly
Perihilar "bat-wing" opacity

Recognition of accompanying signs helps differentiate pulmonary edema from pneumonia.

4. Atelectasis

Air bronchograms may occasionally be present in non-obstructive atelectasis.

However,

complete bronchial obstruction usually eliminates visible air bronchograms because the bronchus itself collapses.

Important clues include:

Volume loss
Fissure displacement
Mediastinal shift
Elevated diaphragm

5. Pulmonary Hemorrhage

Diffuse alveolar hemorrhage produces:

Bilateral consolidation
Ground-glass opacity
Air bronchograms

Clinical clues include:

Hemoptysis
Vasculitis
Autoimmune disease
Anticoagulation

6. Pulmonary Contusion

Following blunt chest trauma:

Patchy consolidation
Air bronchograms
No respect for segmental anatomy

Clinical history is critical.

7. Organizing Pneumonia

Typical findings include

Peripheral consolidation
Reverse halo sign
Air bronchograms

Patients often present with prolonged cough rather than acute sepsis.

8. Invasive Adenocarcinoma (Former Bronchioloalveolar Carcinoma)

Certain lung adenocarcinomas spread along intact alveolar walls.

These tumors may produce:

Persistent consolidation
Air bronchograms
Slowly progressive opacity

Failure of presumed pneumonia to resolve should always raise suspicion for malignancy.

Imaging Pitfalls

Even experienced radiologists occasionally misinterpret air bronchograms.

Several common pitfalls deserve attention.

Pitfall 1

Confusing pulmonary vessels with bronchi.

CT correlation usually resolves uncertainty.

Pitfall 2

Mistaking mucus-filled bronchi for air bronchograms.

A mucus-filled bronchus appears white rather than black.

Pitfall 3

Ignoring patient positioning.

Portable ICU radiographs frequently exaggerate dependent opacities.

Pitfall 4

Overcalling pneumonia.

Diffuse pulmonary edema may closely mimic infectious consolidation.

Clinical context is indispensable.

Pitfall 5

Missing airway obstruction.

If the bronchus is obstructed by tumor or mucus plugging,

true air bronchograms are generally absent.

AI Applications in Air Bronchogram Detection

Chest radiography is one of the most active areas of artificial intelligence in medical imaging. Modern deep learning systems can identify subtle pulmonary abnormalities, prioritize urgent examinations, and assist radiologists in high-volume clinical environments. While early algorithms focused primarily on detecting pneumonia or pleural effusion, contemporary AI platforms increasingly analyze imaging patterns such as consolidation, interstitial changes, and air bronchograms as part of a comprehensive diagnostic assessment.

Rather than replacing radiologists, these systems function as decision-support tools that highlight suspicious regions, quantify disease burden, and integrate imaging findings with electronic health record data. In critically ill patients, especially those with rapidly evolving respiratory failure, this capability can shorten the interval between image acquisition and clinical intervention.

Deep Learning for Chest Radiography

Convolutional neural networks (CNNs) remain the foundation of many chest X-ray AI solutions. Trained on hundreds of thousands of annotated radiographs, these models can detect:

Pulmonary consolidation
Air-space opacity
Pleural effusion
Pneumothorax
Pulmonary edema
Cardiomegaly
Medical devices (e.g., ECMO cannulas, central venous catheters)

When air bronchograms are present within areas of consolidation, AI can assist by identifying the surrounding alveolar disease and directing the radiologist's attention to the affected region. However, final interpretation still depends on clinical context and expert review.

Foundation Models: The Next Evolution of Thoracic Imaging AI

Radiology AI has progressed far beyond detecting isolated abnormalities such as pulmonary nodules or pleural effusions. The newest generation of foundation models is designed to interpret chest imaging in a manner that more closely resembles the reasoning process of experienced radiologists.

Unlike conventional deep learning models trained for a single task, foundation models are pretrained on millions of multimodal data points—including chest radiographs, CT scans, radiology reports, laboratory values, and electronic health records. They learn generalized representations of thoracic disease, enabling them to perform multiple clinical tasks without being retrained for each specific diagnosis.

For patients presenting with respiratory failure, these models can simultaneously:

Detect bilateral consolidation
Identify air bronchograms
Estimate disease severity
Compare with prior examinations
Recognize medical devices such as ECMO cannulas
Suggest differential diagnoses
Generate structured radiology reports

Rather than identifying only one abnormality, foundation models provide a comprehensive overview of the patient's pulmonary status, making them particularly valuable in intensive care settings.

Computer Vision for Air Bronchogram Recognition

Air bronchograms have traditionally required visual recognition by experienced radiologists. However, advances in computer vision now allow AI systems to identify these branching lucencies with remarkable consistency.

A typical computer vision pipeline includes:

Generative AI in Chest Imaging

Generative AI is beginning to transform radiology beyond image detection. Its greatest impact lies in enhancing workflow efficiency and communication.

Potential applications include:

Drafting structured radiology reports from imaging findings
Summarizing longitudinal imaging changes
Comparing current and prior examinations
Translating technical reports into patient-friendly language
Assisting multidisciplinary conferences with concise case summaries
Supporting radiology education by explaining imaging signs such as the air bronchogram

Importantly, generative AI should be viewed as an assistive technology. Final image interpretation, integration of clinical information, and responsibility for patient care remain with the interpreting physician.

Enterprise Clinical AI Platforms

For artificial intelligence to improve patient outcomes, it must integrate seamlessly into the clinical workflow rather than operate as a standalone application.

Modern enterprise AI platforms typically include:

PACS integration
Vendor Neutral Archive (VNA)
Radiology Information System (RIS)
Electronic Health Record (EHR)
HL7 messaging
FHIR interoperability
AI orchestration layer
Structured reporting
Quality assurance dashboard

In this environment, chest radiographs are automatically analyzed immediately after acquisition. AI findings—such as suspected consolidation with air bronchograms—are transmitted directly into the radiologist's worklist, allowing urgent examinations to be prioritized.

This workflow minimizes delays, reduces cognitive burden, and improves consistency across healthcare systems.

Diagnostic Workflow

AI-Enhanced Clinical Workflow for Severe Pneumonia

This workflow mirrors the progression observed in the uploaded case, where imaging findings, microbiological confirmation, and advanced supportive therapy—including ECMO—guided successful management.

Figure Suggestions

Figure 3. Pathophysiology of Air Bronchogram

Figure 4. Enterprise AI Chest Imaging Workflow

Key Imaging Pearls Every Radiologist Should Know
Recognizing an air bronchogram is only the beginning. The true diagnostic value lies in integrating this sign with the patient's clinical presentation, disease
distribution, and associated imaging findings. The following pearls summarize practical lessons from thoracic imaging practice.
1. Air Bronchogram Indicates Alveolar Disease
An air bronchogram almost always signifies that the alveoli have been replaced by fluid, pus, blood, cells, or proteinaceous material while the bronchi remain
patent. It is therefore a hallmark of air-space (alveolar) disease, not an isolated diagnosis.

2. Pneumonia Is the Most Common Cause
When fever, leukocytosis, productive cough, and focal consolidation coexist, bacterial pneumonia should be considered the leading diagnosis.
In the presented case, bronchoalveolar lavage confirmed Moraxella catarrhalis and Haemophilus influenzae infection following the radiographic demonstration of bilateral
air bronchograms.

3. Air Bronchogram Does Not Equal Pneumonia
Although commonly associated with infection, identical imaging findings may occur in:
· Pulmonary edema
· ARDS
· Pulmonary hemorrhage
· Organizing pneumonia
· Pulmonary infarction
· Lung adenocarcinoma with lepidic growth
Always integrate imaging with clinical history and laboratory data.

4. Evaluate Bronchial Patency
A visible air bronchogram implies that the bronchial lumen remains open.
If a bronchus is completely obstructed by a tumor, mucus plug, or foreign body, the classic air bronchogram may disappear. This distinction is valuable when differentiating
obstructive from non-obstructive atelectasis.
5. Assess the Distribution Pattern
The distribution of consolidation provides important diagnostic clues:
· Lobar → Typical bacterial pneumonia
· Multifocal → Bronchopneumonia
· Diffuse bilateral → ARDS or pulmonary edema
· Peripheral → Organizing pneumonia or viral infection
· Dependent → Aspiration or gravity-related edema

6. Portable Chest Radiography Remains Essential
Despite remarkable advances in CT technology, portable chest radiography remains the imaging modality of choice for ICU patients because it enables:
· Bedside evaluation
· Daily monitoring
· ECMO assessment
· Endotracheal tube verification
· Central venous catheter positioning
· Serial assessment of pulmonary consolidation

7. CT Clarifies Equivocal Findings
When chest radiography is inconclusive, CT provides:
· Higher spatial resolution
· Better visualization of bronchial anatomy
· Identification of cavitation
· Detection of pulmonary embolism
· Characterization of diffuse lung disease
· Evaluation of hidden malignancy
8. AI Can Prioritize, Not Replace
Artificial intelligence excels at:
· Detecting pulmonary opacities
· Highlighting suspicious examinations
· Quantifying disease burden
· Comparing serial studies
· Reducing reporting turnaround time
However, AI cannot replace the radiologist's responsibility for integrating imaging findings with patient history, microbiology, and clinical progression.

9. Clinical Context Determines Interpretation
The same chest radiograph may represent entirely different diseases depending on the patient:
· A febrile young adult → Bacterial pneumonia
· A patient with heart failure → Pulmonary edema
· A trauma victim → Pulmonary contusion
· An immunocompromised patient → Opportunistic infection
· A patient with persistent consolidation → Possible malignancy
Pattern recognition must always be accompanied by clinical reasoning.
10. Follow-up Imaging Is Critical
Radiographic improvement often lags behind clinical recovery. Serial imaging helps:
· Confirm therapeutic response
· Detect complications
· Guide ventilator management
· Determine ECMO weaning readiness
· Exclude persistent or recurrent disease
In the featured case, successful treatment allowed discontinuation of ECMO after nine days, followed by liberation from mechanical ventilation and favorable long-term
recovery.

Future Perspectives
The next decade is likely to redefine thoracic imaging through the convergence of artificial intelligence, multimodal data integration, and precision medicine.
Several emerging trends are particularly relevant to chest radiology:
1. Foundation Models as Clinical Copilots
Large multimodal foundation models will increasingly function as intelligent assistants capable of integrating imaging, laboratory results, pathology, genomics, and clinical
notes. Rather than providing isolated predictions, these systems will generate prioritized differential diagnoses and evidence-based recommendations.
2. Continuous ICU Imaging Analytics
Portable chest radiographs acquired in intensive care units may soon be analyzed automatically in real time, enabling continuous monitoring of pulmonary edema, consolidation
progression, pneumothorax, and medical device positioning.
3. Predictive Imaging Biomarkers
Future AI systems will not merely detect disease—they will estimate:
· Risk of respiratory failure
· Probability of mechanical ventilation
· Likelihood of ECMO requirement
· Mortality risk
· Expected hospital length of stay
Such predictive analytics could support earlier intervention and more personalized critical
care.

4. Enterprise-Wide AI Integration
Healthcare organizations are moving toward unified enterprise imaging ecosystems where PACS, RIS, EHR, AI engines, and cloud infrastructure operate as a single
interoperable platform. Standards such as DICOM, HL7, and FHIR will continue to facilitate seamless data exchange, reducing workflow fragmentation and
supporting scalable AI deployment.

5. Explainable and Trustworthy AI
As AI becomes embedded in clinical decision-making, transparency and governance will become increasingly important. Future systems are expected to provide visual
explanations, confidence scores, and traceable reasoning pathways that enable clinicians to understand and verify AI-generated suggestions before acting upon them.

Conclusion
The air bronchogram remains one of the most recognizable and clinically meaningful signs in
thoracic radiology. Its presence indicates preserved bronchial patency within regions of alveolar consolidation and provides a powerful clue to a broad
spectrum of pulmonary disorders, including pneumonia, ARDS, pulmonary edema, hemorrhage, and selected neoplasms.
The featured case illustrates how prompt recognition of this classic imaging sign, combined with microbiological confirmation and aggressive critical care—including ECMO—can
lead to successful recovery even in life-threatening respiratory failure.
Looking ahead, artificial intelligence will increasingly augment—not replace—the expertise of radiologists. Deep learning, foundation models, and enterprise clinical AI
platforms have the potential to enhance detection, streamline workflow, prioritize urgent studies, and support evidence-based decision-making.
Nevertheless, the cornerstone of high-quality radiology will remain the thoughtful integration of imaging findings with clinical context, pathophysiology, and multidisciplinary collaboration.
Ultimately, mastering classic radiologic signs such as the air bronchogram while embracing modern AI technologies represents the optimal path toward safer, faster, and
more precise patient care.
Key Takeaways
· Air bronchogram is one of the most reliable radiographic signs of alveolar consolidation, appearing when air-filled bronchi are
surrounded by fluid- or exudate-filled alveoli.
· In this featured case, recognition of bilateral air bronchograms on a portable chest radiograph facilitated the diagnosis of severe
bacterial pneumonia complicated by acute respiratory distress syndrome (ARDS) and supported timely escalation to ECMO.
· The differential diagnosis of an air bronchogram includes bacterial pneumonia, ARDS, pulmonary edema, diffuse alveolar hemorrhage,
organizing pneumonia, pulmonary contusion, pulmonary infarction, and selected lung adenocarcinomas.
· Chest CT complements chest radiography by confirming bronchial patency, defining the extent of consolidation, and identifying
complications or alternative diagnoses. (The CT image presented in this article is a representative educational example from a different patient.)
· AI-powered chest imaging can automatically detect pulmonary consolidation, prioritize urgent examinations, and assist
radiologists through enterprise PACS and clinical decision support systems. However, AI should augment—not replace—expert clinical interpretation.
· Combining classic radiologic signs, advanced CT correlation, and AI-assisted workflow enables earlier diagnosis,
more efficient triage, and improved outcomes for critically ill patients.
References
1. Hansell DM, Bankier AA, MacMahon H, McLoud TC, Müller NL, Remy J. Fleischner Society: Glossary of Terms for Thoracic Imaging. Radiology.
2008;246(3):697–722. DOI: 10.1148/radiol.2462070712
2. Warren MA, Zhao Z, Koyama T, et al. Severity Scoring of Lung Edema on Chest Radiography Is Associated With Clinical Outcomes in
ARDS. Thorax. 2018;73:840–846. DOI: 10.1136/thoraxjnl-2017-211280
3. Rubin GD. COVID-19 Pneumonia and the Value of Chest Imaging. New England Journal of Medicine. 2020;382:1463–1464. DOI: 10.1056/NEJMe2007273
4. Topol EJ. High-performance Medicine: The Convergence of Human and Artificial Intelligence. Nature Medicine.
2019;25:44–56. DOI: 10.1038/s41591-018-0300-7
5. Rajpurkar P, Irvin J, Zhu K, et al. CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning. arXiv.
2017. (Foundational AI work; no DOI assigned.)
6. Esteva A, Robicquet A, Ramsundar B, et al. A Guide to Deep Learning in Healthcare. Nature Medicine. 2019;25:24–29. DOI: 10.1038/s41591-018-0316-z
7. McKinney SM, et al. International Evaluation of an AI System for Breast Cancer Screening. Nature. 2020;577:89–94. DOI: 10.1038/s41586-019-1799-6
8. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key Challenges for Delivering Clinical Impact with Artificial
Intelligence. BMC Medicine. 2019;17:195. DOI: 10.1186/s12916-019-1426-2
9. Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL. Artificial Intelligence in Radiology. Nature Reviews Cancer.
2018;18:500–510. DOI: 10.1038/s41568-018-0016-5
10. The featured educational image case: Air Bronchogram. New England Journal of Medicine Images in Clinical Medicine. DOI: 10.1056/NEJMicm1503806

Air Bronchogram Explained: Imaging Findings, Differential Diagnosis, and AI in Modern Chest Radiology