Deep learning for chest X-ray analysis: A survey

Medical Image Analysis, Journal Year: 2021, Volume and Issue: 72, P. 102125 - 102125

Published: June 5, 2021

Recent advances in deep learning have led to a promising performance many medical image analysis tasks. As the most commonly performed radiological exam, chest radiographs are particularly important modality for which variety of applications been researched. The release multiple, large, publicly available X-ray datasets recent years has encouraged research interest and boosted number publications. In this paper, we review all studies using on published before March 2021, categorizing works by task: image-level prediction (classification regression), segmentation, localization, generation domain adaptation. Detailed descriptions included commercial systems field described. A comprehensive discussion current state art is provided, including caveats use public datasets, requirements clinically useful gaps literature.

Language: Английский

---

The need to separate the wheat from the chaff in medical informatics

International Journal of Medical Informatics, Journal Year: 2021, Volume and Issue: 153, P. 104510 - 104510

Published: June 2, 2021

Language: Английский

---

How does artificial intelligence in radiology improve efficiency and health outcomes?

Pediatric Radiology, Journal Year: 2021, Volume and Issue: 52(11), P. 2087 - 2093

Published: June 12, 2021

Abstract Since the introduction of artificial intelligence (AI) in radiology, promise has been that it will improve health care and reduce costs. Has AI able to fulfill promise? We describe six clinical objectives can be supported by AI: a more efficient workflow, shortened reading time, reduction dose contrast agents, earlier detection disease, improved diagnostic accuracy personalized diagnostics. provide examples use cases including available scientific evidence for its impact based on hierarchical model efficacy. conclude market is still maturing little known about contribution practice. More real-world monitoring practice expected aid determining value making informed decisions development, procurement reimbursement.

Language: Английский

---

Value-added Opportunistic CT Screening: State of the Art

Radiology, Journal Year: 2022, Volume and Issue: 303(2), P. 241 - 254

Published: March 15, 2022

Opportunistic CT screening leverages robust imaging data embedded within abdominal and thoracic scans that are generally unrelated to the specific clinical indication have heretofore gone largely unused. This incidental information may prove beneficial patients in terms of wellness, prevention, risk profiling, presymptomatic detection relevant disease. The growing interest CT-based opportunistic relates a confluence factors: objective generalizable nature body composition measures, emergence fully automated explainable AI solutions, sheer volume performed, increasing emphasis on precision medicine value-added initiatives. With systematic approach other useful markers, initial evidence suggests their ability help radiologists assess biologic age predict future adverse cardiometabolic events rivals even best available reference standards. Emerging suggest standalone “intended” over an unorganized be justified, especially when combined with established cancer screening. review will discuss current status screening, including markers various disease processes impacted. remaining hurdles widespread adoption include generalization more diverse patient populations, disparate technical settings, reimbursement. © RSNA, 2022 An earlier incorrect version appeared online. article was corrected March 18, 2022.

Language: Английский

---

Robust and data-efficient generalization of self-supervised machine learning for diagnostic imaging

Nature Biomedical Engineering, Journal Year: 2023, Volume and Issue: 7(6), P. 756 - 779

Published: June 8, 2023

Language: Английский

---

Computer vision in surgery: from potential to clinical value

npj Digital Medicine, Journal Year: 2022, Volume and Issue: 5(1)

Published: Oct. 28, 2022

Abstract Hundreds of millions operations are performed worldwide each year, and the rising uptake in minimally invasive surgery has enabled fiber optic cameras robots to become both important tools conduct sensors from which capture information about surgery. Computer vision (CV), application algorithms analyze interpret visual data, a critical technology through study intraoperative phase care with goals augmenting surgeons’ decision-making processes, supporting safer surgery, expanding access surgical care. While much work been on potential use cases, there currently no CV widely used for diagnostic or therapeutic applications Using laparoscopic cholecystectomy as an example, we reviewed current techniques that have applied their clinical applications. Finally, discuss challenges obstacles remain be overcome broader implementation adoption

Language: Английский

---

Implementation of Clinical Artificial Intelligence in Radiology: Who Decides and How?

Radiology, Journal Year: 2022, Volume and Issue: 305(3), P. 555 - 563

Published: Aug. 2, 2022

As the role of artificial intelligence (AI) in clinical practice evolves, governance structures oversee implementation, maintenance, and monitoring AI algorithms to enhance quality, manage resources, ensure patient safety. In this article, a framework is established for infrastructure required implementation presents road map governance. The answers four key questions: Who decides which tools implement? What factors should be considered when assessing an application implementation? How applications implemented practice? Finally, how monitored maintained after Among many challenges practice, devising flexible that can quickly adapt changing environment will essential quality care improvement objectives. © RSNA, 2022 An earlier incorrect version appeared online. This article was corrected on August 2, 2022.

Language: Английский

---

Comparative Performance of ChatGPT and Bard in a Text-Based Radiology Knowledge Assessment

Canadian Association of Radiologists Journal, Journal Year: 2023, Volume and Issue: 75(2), P. 344 - 350

Published: Aug. 14, 2023

Purpose Bard by Google, a direct competitor to ChatGPT, was recently released. Understanding the relative performance of these different chatbots can provide important insight into their strengths and weaknesses as well which roles they are most suited fill. In this project, we aimed compare recent version ChatGPT-4, in ability accurately respond radiology board examination practice questions. Methods Text-based questions were collected from 2017-2021 American College Radiology’s Diagnostic Radiology In-Training (DXIT) examinations. ChatGPT-4 queried, comparative accuracies, response lengths, times documented. Subspecialty-specific analyzed well. Results 318 included our analysis. ChatGPT answered significantly more than (87.11% vs 70.44%, P < .0001). ChatGPT’s length shorter Bard’s (935.28 ± 440.88 characters 1437.52 415.91 characters, time longer (26.79 3.27 seconds 7.55 1.88 seconds, performed superiorly neuroradiology, (100.00% 86.21%, = .03), general & physics (85.39% 68.54%, .001), nuclear medicine (80.00% 56.67%, .01), pediatric (93.75% 68.75%, ultrasound 63.64%, .001). remaining subspecialties, there no significant differences between performance. Conclusion displayed superior knowledge compared Bard. While both display reasonable knowledge, should be used with conscious limitations fallibility. Both provided incorrect or illogical answer explanations did not always address educational content question.

Language: Английский

---

Autonomous Chest Radiograph Reporting Using AI: Estimation of Clinical Impact

Radiology, Journal Year: 2023, Volume and Issue: 307(3)

Published: March 7, 2023

Background Automated interpretation of normal chest radiographs could alleviate the workload radiologists. However, performance such an artificial intelligence (AI) tool compared with clinical radiology reports has not been established. Purpose To perform external evaluation a commercially available AI for (a) number autonomously reported, (b) sensitivity detection abnormal radiographs, and (c) that reports. Materials Methods In this retrospective study, consecutive posteroanterior from adult patients in four hospitals capital region Denmark were obtained January 2020, including images emergency department patients, in-hospital outpatients. Three thoracic radiologists labeled reference standard based on radiograph findings into following categories: critical, other remarkable, unremarkable, or (no abnormalities). classified as high confidence (normal) (abnormal). Results A total 1529 included analysis (median age, 69 years [IQR, 55–69 years]; 776 women), 1100 (72%) by having 617 (40%) critical 429 (28%) radiographs. For comparison, text insufficient excluded (n = 22). The was 99.1% (95% CI: 98.3, 99.6; 1090 patients) 99.8% 99.1, 99.9; 616 Corresponding sensitivities radiologist 72.3% 69.5, 74.9; 779 1078 93.5% 91.2, 95.3; 558 597 patients), respectively. Specificity AI, hence potential autonomous reporting rate, 28.0% all 23.8, 32.5; 120 7.8% (120 Conclusion Of 28% reported any abnormalities higher than 99%. This corresponded to entire production. © RSNA, 2023 Supplemental material is article. See also editorial Park issue.

Language: Английский

---

ChatGPT and Generating a Differential Diagnosis Early in an Emergency Department Presentation

Annals of Emergency Medicine, Journal Year: 2023, Volume and Issue: 83(1), P. 83 - 86

Published: Sept. 9, 2023

Language: Английский

---