Skip to content

Review: Recent Developments and Future Trends in Thoracic Radiology

Key findings

  • This review discusses recent advances in artificial intelligence for thoracic imaging, new techniques that will create new applications and challenges in training and recruiting thoracic radiologists
  • Artificial intelligence is already used in automated or assisted detection algorithms and tools that facilitate radiologist performance
  • Future applications of artificial intelligence will improve triage, characterization of complex conditions and measurement of biomarkers, and allow routine integration of quantitative data into reporting of chest CT
  • New and developing imaging techniques include photon-counting detector CT, dynamic cine MRI, hyperpolarized xenon MRI, and dark-field chest radiography

Advances in CT, MRI and PET have resulted not only in qualitative improvement of chest imaging but also the ability to provide functional data and enable quantitative assessment.

Theresa C. McLoud, MD, senior advisor for faculty affairs in the Radiology Department at Massachusetts General Hospital and professor of Radiology at Harvard Medical School, and Brent P. Little, MD, of Mayo Clinic Florida, reviewed in Radiology the recent advances in thoracic imaging that are expected to improve patient management.

Artificial Intelligence (AI) and Quantitative Imaging

Automated or assisted detection algorithms—Certain AI algorithms are already FDA-approved for detection of line and tube positions, nodules, pulmonary emboli, pneumothorax and rib fractures. Radiologists can expect steady improvements in the breadth and accuracy of using AI as a second reader, as studies are showing greater reader confidence and reduced reading times.

Applications for streamlining time-intensive processes—AI applications currently allow automated quantification of coronary arterial calcification, emphysema, and body mass composition. Tools for automated volumetric measurements and grading the severity of parenchymal lung diseases are under development. Routine use of such algorithms may save radiologists time and standardize quantitative reporting of CT biomarkers.

Triage—Algorithms that flag emergent findings such as pulmonary emboli or pneumothorax are likely to be made available.

Characterizing complex conditions—AI algorithms are promising for identifying, characterizing, severity grading and clinical prognostication of cancer, pneumonia, interstitial lung disease, and other complex pathologies. Many algorithms designed to detect pulmonary tuberculosis are already being used.

"Opportunistic screening" is becoming more frequent. This term refers to integrating quantitative data into routine reporting of chest CT, with measurements of disease severity and automated comparison with previous examinations.

Quantification of biomarkers—AI could allow routine measurement of biomarkers such as coronary arterial calcification, aortic diameter and calcification, body mass composition, cardiac chamber measurements, bone density, lung density, lung volume and airway measurements, with comparison to population-based normal ranges.

Radiomics, the use of imaging features extracted by computer algorithms to predict characteristics of disease, shows promise for evaluating cancer and interstitial lung disease.

Hardware Innovations

New and emerging imaging techniques will continue to refine patient care:

  • Photon-counting detector CT uses semiconductors that convert X-ray photons directly to electronic pulses, improving spatial resolution, and reducing noise and radiation dose
  • Advances in airway MRI include the use of dynamic cine MRI to evaluate tracheomalacia and the use of hyperpolarized xenon MRI, which provides functional mapping and quantification of pulmonary perfusion, ventilation and gas exchange, and may be particularly useful for assessing patients with long COVID
  • Dark-field chest radiography, another technique for functional analysis, can boost the signal in diseases such as fibrosis, emphysema and COVID-19 pneumonia

The review also discusses the potential for using biomarkers to decrease the very high rate of false-positive rate with low-dose CT for lung cancer screening, and changes in the training and recruitment of thoracic radiologists.

Learn more about the Division of Thoracic Imaging and Intervention

Learn more about the Department of Radiology

Related topics


Based on six-reader review of 150 CT examinations, Brent P. Little, MD, of the Department of Radiology, and colleagues urge their fellow radiologists to recognize that CT features of COVID-19 pneumonia overlap not just with influenza pneumonia but also non-infectious organizing pneumonia.


Massachusetts General Hospital researchers have devised an artificial intelligence system for automatic labeling of chest X-rays in which the user can specify a tradeoff between accuracy and efficiency. This innovation might reduce the need for labor-intensive manual labeling of very large datasets by human experts.