AJR Podcasts

We all make financial mistakes. Cohosts Sherry S. Wang, MBBS, and Surbhi Raichandani, MD, discuss their own mistakes, what they have learned, and how they are now doing things differently. Listen to their discussion in episode 5 of Difficult Conversations, an AJR Podcast Series.
https://www.ajronline.org/doi/10.2214/AJR.26.36138

Generative artificial intelligence tools are being increasingly used in radiology reporting. Alan Gao, MD, discusses this article by Dogra et al. that explores the role of supervised fine-tuning corpus composition in impacting management-guidance behavior in these tools' generated impressions.
Full article: Training Corpus Composition Independently Affects Positive and Negative Guidance in Artificial Intelligence–Generated Chest CT Impressions: A Dose-Response Study

When community practice meets the academic enterprise, opportunities and tensions can emerge in unexpected ways. Alan Woronoff, MD, speaks with cohosts Winnie Hahn, MD, and Elizabeth Hecht, MD, about his transition from private practice to an academically affiliated health system, exploring inclusion, influence, community versus academic priorities, subspecialization, productivity, professional growth, and lessons for radiologists and leaders. Listen to their discussion in episode 4 of Mentorship Unfiltered, an AJR Podcast Series.

Clinical data drive meaningful research, but access comes with responsibility. Andrea Doria, MD PhD, speaks with host Amit Gupta, MD, about navigating research ethics, informed consent, deidentification, dataset bias, responsible data sharing, and patient trust. Listen to their discussion in episode 4 of The Early-Career Researcher's Playbook, an AJR Podcast Series.
*Key Takeaways
Ethics as Good Science: Ethical guidelines are not administrative obstacles. They are an integral part of responsible data stewardship and maintaining the trust patients place in the healthcare system.
The Consent Misconception: Retrospective studies do not automatically bypass patient consent. Only the ethics board can grant a waiver of consent, for example for scenarios with large unreachable patient populations.
Anonymization vs. De-identification: Removing a name is merely anonymization. True de-identification requires scrubbing private DICOM metadata, redacting burned-in pixels on ultrasounds, and accounting for 3D reconstructions that might reveal unique physical patient characteristics.
External Data Sharing: Before uploading clinical data to external cloud platforms or AI tools, institutions should establish strict Data Transfer Agreements detailing storage protocols, access rights, and data deletion timelines.
Scope Creep and Amendments: Reusing an existing dataset to answer a new research question unapproved by the IRB is an ethical violation. Researchers must submit an amendment or a new application to the ethics board before shifting the scope of their study.
*Chapters
0:00 - Welcome
02:21 - When Ethics Approval Starts
04:40 - Consent and Waivers
09:26 - Deidentifying Imaging Data
12:48 - Sharing Data Externally
15:31 - Bias and Representation
18:41 - Stewardship Beyond Rules
21:30 - Reusing Data and Amendments
26:34 - Quick Fire Playbook
33:23 - Final Takeaways
Follow AJR on Social Media
LinkedIn: https://www.linkedin.com/showcase/ajr-radiology/
YouTube: https://www.youtube.com/channel/UCfFAYezkLMxJGMgIJLN0Dpg
Instagram: https://www.instagram.com/ajr_radiology/
TikTok: https://www.tiktok.com/@ajr_radiology
X: https://x.com/AJR_Radiology
BlueSky: https://bsky.app/profile/ajrradiology.bsky.social
Threads: https://www.threads.com/@ajr_radiology
*These sections were generated using artificial intelligence (Descript and Google Gemini) and then reviewed for accuracy.

Unsuccessful FFRCT analysis can be influenced by the underlying CT technology used. Radhika Rajeev, MD, discusses the article by Balasubramanian et al. exploring causes of such rejected analyses for two CT scanners.
Full article: FFRCT Analysis Rejections: Comparison of Frequency and Associated Factors Between Two CT Scanners