Akshatha Mohan

I build clinically grounded AI for medical imaging.

I work in the Medical AI Lab at Washington University in St. Louis (MIR/CIRC), led by Dr. Shinjini Kundu, MD, PhD. My research focuses on robust imaging pipelines, task-driven evaluation, and translational medical AI systems.

In Duke collaborative work, I contributed to a task-driven evaluation of 13 image quality metrics across NLST, VLST, and DLCS datasets for lung nodule detection, lesion-centric cancer classification, and COPD quantification with CT harmonization.

Curriculum Vitae (PDF)

News

  • 2026: Continuing medical imaging AI research in MIR/CIRC with focus on clinically meaningful evaluation.
  • 2025: Contributed to Duke collaborative manuscript on task-driven IQA and CT harmonization analysis.
  • 2024: Publications across CVPRW, ICMLA, and SPIE on texture analysis and interpretable computer vision.

Impact Summary

  • Integrated structural, perceptual, and statistical IQA perspectives to evaluate cross-cohort CT reliability.
  • Connected image quality behavior directly to downstream clinical tasks instead of relying on generic visual metrics alone.
  • Built end-to-end AI systems, including retrieval-augmented LLM applications for practical decision-support workflows.

Selected Publications

Manuscript in Progress Lung CT IQA manuscript visual

Task-Driven Evaluation of Image Quality Metrics for Lung CT AI Workflows

A. Mohan, collaborators

Duke collaboration spanning NLST, VLST, and DLCS analyses for detection, classification, and harmonization tasks.

CVPRW 2024 Lacunarity pooling visual

Lacunarity Pooling Layers for Plant Image Classification Using Texture Analysis

A. Mohan, J. Peeples

Paper link