Deep Learning Based Detection and Invasion-Oriented Stage Classification of Colorectal Cancer

Colorectal cancer prognosis depends on how deeply the tumour invades surrounding tissue, yet invasionoriented staging has received little attention in deep learning studies.This paper presents a two-stage framework for histopathological patches and whole slide images.A binary classifier first separates normal from tumour tissue, and tumour patches are then assigned one of four invasion-oriented stages, Stage 0 to Stage 3.Both models use a hybrid EfficientNet-B0 and Transformer network with cross attention fusion, trained on STARC-9 patches with Macenko stain normalisation.On held-out patches, the binary and stage models reached accuracies of 99.64 and 99.70 percent.Patch predictions from TCGA-COAD and TCGA-READ slides are aggregated into slide reports; on 160 labelled slides, slide-level stage accuracy was 83.75 percent, showing a clear gap from patch-level results.A Flask web interface exposes the pipeline.

Authors

Publication Details

Journal
IJARCCE
Published
2026-11-18
DOI
https://doi.org/10.17148/ijarcce.2026.1511153
Primary Topic
AI in cancer detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Deep Learning Based Detection and Invasion-Oriented Stage Classification of Colorectal Cancer

Sinchana S, Shrthika A, Sudeep P, Sharath Kumar B et al.
1 links
IJARCCE
AI in cancer detection
article

Deep Learning Based Detection and Invasion-Oriented Stage Classification of Colorectal Cancer

Sinchana S, Shrthika A, Sudeep P, Sharath Kumar B, Pramod Kumar P M
article en

Abstract

Colorectal cancer prognosis depends on how deeply the tumour invades surrounding tissue, yet invasionoriented staging has received little attention in deep learning studies.This paper presents a two-stage framework for histopathological patches and whole slide images.A binary classifier first separates normal from tumour tissue, and tumour patches are then assigned one of four invasion-oriented stages, Stage 0 to Stage 3.Both models use a hybrid EfficientNet-B0 and Transformer network with cross attention fusion, trained on STARC-9 patches with Macenko stain normalisation.On held-out patches, the binary and stage models reached accuracies of 99.64 and 99.70 percent.Patch predictions from TCGA-COAD and TCGA-READ slides are aggregated into slide reports; on 160 labelled slides, slide-level stage accuracy was 83.75 percent, showing a clear gap from patch-level results.A Flask web interface exposes the pipeline.

IJARCCEVol. 15(11)
Good health and well-being
AI in cancer detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon