Upstream Machine Learning in Radiology

Sandino, Christopher M.; Cole, Elizabeth K.; Alkan, Cagan; Chaudhari, Akshay; Loening, Andreas M.; Hyun, Dongwoon; Dahl, Jeremy J.; Imran, Abdullah-Al-Zubaer; Wang, Adam; Vasanawala, Shreyas S.

doi:10.1016/j.rcl.2021.07.009

Cited by 10 publications

(2 citation statements)

References 75 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Emergency MRI is currently limited in terms of availability but might become more feasible and effective in the future due to technological advances. Possible developments for the near future include improved hardware, such as low field [ 41 ] and portable [ 42 ] MRI, as well as faster scanning using acceleration techniques [ 43 ], artificial intelligence-based reconstruction [ 44 ], and synthetic MRI sequences [ 45 ]. Thus, wider adoption of emergency MRI for acute neck infections can reasonably be foreseen.…”

Section: Future Directionsmentioning

confidence: 99%

MRI of acute neck infections: evidence summary and pictorial review

et al. 2023

View full text Add to dashboard Cite

Infection of the deep neck spaces is a life-threatening acute illness that requires prompt diagnosis and treatment. Magnetic resonance imaging (MRI) offers unsurpassed soft tissue discrimination and is therefore well suited for imaging neck infections. Recently, the feasibility, diagnostic accuracy, and clinical significance of this method have been documented in patients with acute neck infections. This review article summarizes the scientific evidence, provides a practical guide to image acquisition and interpretation, reviews the most common imaging findings, and discusses some difficult diagnoses and pitfalls in acute neck infections, to help both radiologists and clinicians in managing these critically ill patients.

show abstract

Section: Future Directionsmentioning

confidence: 99%

MRI of acute neck infections: evidence summary and pictorial review

et al. 2023

View full text Add to dashboard Cite

show abstract

“…However, these two works are limited to musculoskeletal MRI protocols. Few AI systems have been developed for protocol, contrast-agent, and machine-specification selection [10], [11], [12].…”

Section: Introductionmentioning

confidence: 99%

Radiology Decision Support System for Selecting Appropriate CT Imaging Titles Using Machine Learning Techniques Based on Electronic Medical Records

Shokrollahi,

Chaves,

Lam

et al. 2023

IEEE Access

Self Cite

View full text Add to dashboard Cite

Radiologists use an imaging order from the ordering physician, which includes a radiology title, to select the most suitable imaging protocol. Inappropriate radiology titles can disrupt protocol selection and result in mistaken or delayed diagnosis. The objective of this work is to develop an algorithm to predict correct radiology titles from incoming exam order data. The proposed instrument is an ensemble of five decision tree-based machine learning (ML) techniques (Light Gradient Boosting Machine, eXtreme Gradient Boosting Machine, Random Forest, Adaptive Boosting, and Random UnderSampling Boosting Model) trained to recommend radiology titles of computed tomography imaging examinations based on electronic medical records. Issues of imbalanced data and generalization were addressed. The tuned models were used to predict the top three radiology titles for the radiologist revision. The models were evaluated using a 10-fold cross-validation method, yielding an approximate average accuracy of 80.5% ± 2.02% and F1-score of 80.3% ± 1.67% for all models, while the ensemble classifier (∼83% F1-score) outperformed individual models. An accumulated average accuracy of ∼92% was obtained for the top three predictions. ML techniques can predict radiology titles and identify highly important features. The proposed system can guide physicians toward selecting appropriate radiology titles and alert radiologists to inconsistencies between the radiology title in the exam order and the patient's underlying conditions, thereby improving imaging utility and increasing diagnostic accuracy, which favors better patient outcomes.

show abstract