Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Explainability and Trust in Deep Learning for Cancer Imaging: Systematic Barriers, Clinical Misalignment, and a

Surekha Borra¹, Nilanjan Dey², Simon Fong³

¹Department of CSE (ICB), K.S. Institute of Technology, Bangalore 560109, India.

|May 13, 2026

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

One probe, two chemistries: an orthogonal fluorescent sensing platform for glutathione and hydrazine in biological fluids and food samples.

Journal of materials chemistry. B·2026

Same author

Decoding structural rigidity and charge-transfer polarization in barbituric-acid-based donor-π-acceptor chromophores.

RSC advances·2026

Same author

Interplay of Electronic Delocalization and Aggregation Dynamics in Decoding Metabolic Disease Biomarkers: Application to Screening of Biological Fluids and Intracellular Imaging.

Langmuir : the ACS journal of surfaces and colloids·2026

Same author

RayFlex: Inducing Weight Perception Through Raycast Pseudo-Haptics in Virtual Reality.

IEEE transactions on visualization and computer graphics·2026

Same author

Clinical feasibility of intratracheal tracheostomy sealing using a novel sealing disc prototype.

Scientific reports·2026

Same author

Bisbenzimidazoles with a benzenediyl spacer: an efficient fluorophore for dihydrogen phosphate sensing and targeted living cell imaging.

Analytical methods : advancing methods and applications·2026

Explainable AI (XAI) in oncology imaging faces trust barriers like bias and opacity. Durable trust requires integrating explainability as a core design principle for dependable clinical AI.

Area of Science:

Artificial Intelligence
Medical Imaging
Oncology

Background:

Deep learning (DL) excels in cancer imaging but lacks explainability and reliable uncertainty estimation, hindering clinical adoption.
Explainable AI (XAI) aims to address these limitations, but challenges like dataset bias and model opacity persist.
Trust in AI systems is crucial for clinical integration in oncology.

Purpose of the Study:

To review the challenges, impact, and translational implications of XAI in oncology imaging.
To identify barriers to trust in explainable AI models for cancer diagnostics.
To outline a pathway for developing accountable and clinically dependable AI systems in oncology.

Main Methods:

Review of research questions examining XAI in oncology imaging.

Keywords:

algorithmic bias cancer imaging clinical trust deep learning explainable artificial intelligence human–AI collaboration regulatory governance uncertainty calibration

Related Experiment Videos

Identification of key barriers to trust, including dataset bias, shortcut learning, and model opacity.

Evaluation of architectural and post hoc techniques for enhancing model interpretability.

Main Results:

Explainable models can improve clinician confidence and decision-making if explanations are faithful, meaningful, and uncertainty-aware.
Barriers to trust include dataset bias, shortcut learning, CNN opacity, and workflow misalignment.
Explainability alone is insufficient; durable trust requires epistemic alignment, validation, governance, and equity evaluation.

Conclusions:

Reframing explainability as a structural design principle is key for clinical AI.
Achieving accountable and dependable AI in oncology requires a holistic approach beyond performance metrics.
Future AI development must prioritize trust, clinical alignment, and ethical considerations for successful integration.