Beyond Knowledge Deficits: Cognitive Framing Errors in Radiology

Institute for Responsible Healthcare AI

Summary Core Theory Interactive Tool References

White Paper / Educational Module

Beyond Knowledge Deficits: Cognitive Framing Errors in Radiology

From Visual Impression to Morphologic Reasoning

SM

Dr. Sharad Maheshwari, MD

Founder, Institute for Responsible Healthcare AI

Executive Summary

Educator's Insight

We propose the concept of the Cognitive Framing Deficit: a phenomenon where trainees bypass structured morphologic analysis, moving directly from visual impression to semantic labeling. The resulting terminology is technically inaccurate, even when detection is flawless.

1. The Diagnostic Error Paradigm

Medical education traditionally frames diagnostic errors through three major lenses: Knowledge deficits, Perceptual misses, and Experience limitations. Yet, these fail to explain a rampant phenomenon: The resident correctly identifies an abnormality but inaccurately characterizes its morphology.

• Large tumors equated with aggressive growth patterns automatically.

2. The Cognitive Framing Deficit

A Cognitive Framing Deficit occurs when a trainee assigns a diagnostic descriptor before completing a systematic morphologic analysis. The omission of intermediate analytical steps creates opportunities for flawed reasoning.

3. The Infiltrative-Lobulated Paradox

Consider a large heterogeneous hepatic mass. An experienced radiologist interrogates the image for interfaces, mass effect, and vessel interaction. A resident may instead observe large size, heterogeneity, and an aggressive appearance, immediately concluding the lesion is "infiltrative."

This is a cognitive shortcut. The conclusion is reached before morphology has been systematically analyzed.

4. Cognitive Psychology of Reporting Errors

4.1 Prototype Matching

Residents often develop simplified mental models: "Infiltrative tumors are large, heterogeneous, and irregular." When a lesion resembles this prototype, the label is automatically applied. However, heterogeneity is not infiltration.

4.2 Feature Substitution

A concept described by Kahneman. The difficult question ("Is this lesion infiltrative?") is unconsciously replaced by an easier one ("Does this look aggressive?"). The answer to the easier question replaces the harder one.

4.3 Gestalt Dominance

Novices process global impressions (Large, Necrotic) before local details (Margins, Interfaces), assigning aggressive labels globally before verifying locally.

5. Uncertainty Management and Diagnostic Hedging

Residents frequently encounter lesions they know are abnormal but cannot confidently classify. The resulting discomfort creates pressure to produce a reportable description. Terms such as Infiltrative, Concerning, or Ill-defined often function as linguistic safety mechanisms.

6. Reverse Diagnostic Reasoning

Instead of allowing morphology to dictate behavior and ultimately diagnosis, trainees often perform a "top-down" error:

Possible Diagnosis (Cholangiocarcinoma)

Expected Morphology (Infiltrates)

Observed Morphology (Falsely labeled "Infiltrative")

7. The Morphology-Language Disconnect Matrix

8-10. Educational Intervention: The "Show Me Where It Stops" Principle

When a trainee states, "This lesion is infiltrative," the expert response should be: "Show me where the tumor stops." This simple question shifts the discussion from gestalt impression to evidence-based interface analysis.

Interactive Tool

Resident Cognitive Discipline Calculator

Input the objective morphologic findings below to assess the appropriate terminology and detect potential Cognitive Framing Errors.

1. Margin/Interface -- Select -- Well-defined / Smooth Poorly-defined / Irregular

2. Vessel Interaction -- Select -- Vessels Displaced Vessels Encased / Invaded

3. Overall Size/Heterogeneity -- Select -- Small / Homogeneous Large / Heterogeneous

4. Trainee's Proposed Term -- Select -- Expansile Infiltrative

11-12. Implications & Future AI Integration

Training programs must explicitly teach morphologic reasoning, diagnostic language, cognitive biases, and uncertainty management. The goal is disciplined characterization, not merely lesion detection.

Future AI-assisted reporting systems can enforce this cognitive sequence by barring immediate top-level diagnostic labels until the underlying structural parameters (Contour → Interface → Architecture → Growth Pattern) are populated.

"The central challenge in advanced radiology education is not knowledge acquisition. It is cognitive discipline."

References & Suggested Reading

• [1]
  Kahneman D. Thinking, Fast and Slow. Farrar, Straus and Giroux; 2011. (Concept of Feature Substitution and Heuristics).
• [2]
  Busby LP, Courtier JL, Gleason AW. Bias in Radiology: The How and Why of Misses and Misinterpretations. Radiographics. 2018;38(1):236-247. (PubMed ID: 29320328)
• [3]
  Norman GR, Eva KW. Diagnostic error and clinical reasoning. Med Educ. 2010;44(1):94-100. (PubMed ID: 20078760)
• [4]
  Berlin L. Radiologic errors and malpractice: a blurry distinction. AJR Am J Roentgenol. 2007;189(3):517-522. (PubMed ID: 17715094)

© 2026 Institute for Responsible Healthcare AI