Life of Research
Upload
Life of Research
Sign inCreate account
Appearance

Life of Research · open-access repository · LOR-00159

Journal article Open access CC BY 4.0

Two-Stage Transfer Learning for Slice-Level Pancreatic CT Classification: Performance Benchmarking, Confidence-Interval Reporting, Saliency Audit and a Data-Leakage Risk Analysis

  • vemulawada Navyasri

Journal article

LOR-00159

Permanent link: https://lifeofresearch.com/articles/159

Download PDF (1015.5 KB) Cite

Cite

Two-Stage Transfer Learning for Slice-Level Pancreatic CT Classification: Performance Benchmarking, Confidence-Interval Reporting…

 Read

Abstract

Transfer learning with partial fine-tuning separates the two classes almost perfectly on a slice-level split, and fine-tuning EfficientNet-B0 removes the baseline's specificity failure. However, the reported figures should be read as an upper bound on clinical performance until they are reproduced with patient-level partitioning and external data. We provide a reproducible leakage-aware protocol and a reporting checklist for slice-based pancreatic CT classifiers.

Keywords

pancreatic cancer; computed tomography; deep learning; transfer learning; EfficientNet; Grad-CAM; data leakage; medical image classification

Files (1)

Cite this record

Navyasri, V. (2026). Two-Stage Transfer Learning for Slice-Level Pancreatic CT Classification: Performance Benchmarking, Confidence-Interval Reporting, Saliency Audit and a Data-Leakage Risk Analysis. Life of Research. https://lifeofresearch.com/articles/159

APA 7th edition
Export to a reference manager
BibTeX 
@misc{navyasri2026two,
  author       = {Navyasri, vemulawada},
  title        = {{Two-Stage Transfer Learning for Slice-Level Pancreatic CT Classification: Performance Benchmarking, Confidence-Interval Reporting, Saliency Audit and a Data-Leakage Risk Analysis}},
  year         = {2026},
  month        = oct,
  date         = {2026-10-30},
  publisher    = {Life of Research},
  url          = {https://lifeofresearch.com/articles/159},
  abstract     = {Transfer learning with partial fine-tuning separates the two classes almost perfectly on a slice-level split, and fine-tuning EfficientNet-B0 removes the baseline's specificity failure. However, the reported figures should be read as an upper bound on clinical performance until they are reproduced with patient-level partitioning and external data. We provide a reproducible leakage-aware protocol and a reporting checklist for slice-based pancreatic CT classifiers.},
  keywords     = {pancreatic cancer; computed tomography; deep learning; transfer learning; EfficientNet; Grad-CAM; data leakage; medical image classification},
  language     = {en},
  copyright    = {Creative Commons Attribution 4.0}
}

Using Zotero or Mendeley? Their browser connector saves this record from this page, with its PDF when there is one. On search and browse pages it lists every result so you can save many at once.

Guides What a complete citation needs What CC BY 4.0 allows

  1. A Rewrite That Keeps the Label Has Already Read It: Why an LLM Paraphrase Can Raise a Sentiment Classifier's Accuracy Without the Classifier Getting Better, Four Accounts of the Gain, and the Re-Annotation and Neutral-Rate Controls That Would Separate Them Pranay Mahendrakar · 2026 · Journal article
  2. Compressed Once, Read Many Times: Why Prompt-Compression Results Do Not Transfer to Agent Memory, What the Query-Agnostic Line Already Showed, and the Two Properties of a Memory Write No Protocol Yet Scores Pranay Mahendrakar · 2026 · Journal article
  3. A Physics-Constrained Decision Framework for Orbital-Debris Intervention Search Mahib Sadman · 2026 · Journal article
  4. A Lesson Is an Untested Counterfactual: The Three Claims Inside a Stored Agent Failure Memory, Which of Them Anything Checks, and Why Every Published Retention Rule Keys on a Proxy Pranay Mahendrakar · 2026 · Journal article
  5. Interpreting Statistical Results for High-Quality Research Reporting ROGESH · 2026 · Journal article

More related research →