All researchProject overview

Medical imaging · 2026

Kidney stone segmentation in CT

Finding the CT slices most likely to contain kidney stones before tracing their boundaries.

Qatar University MSc research · Qatar University Student Grant QUST-1-CENG-2023-795 and Medical Research Council grant MRC01-21-1027

Pipeline diagram. Abdominal CT volumes are preprocessed, a classifier keeps the slices that contain kidney, a second classifier keeps the slices that contain stones, and a segmentation model outlines the kidney and stones on the remaining slices.
Figure 1 from Mahmud et al. (2026). Preprocessed CT slices pass through two successive classifiers, which keep kidney slices and then stone slices, before KSSNet segments the remaining slices. Green arrows mark features transferred between the stages.View full-size figureSource article© 2026 Elsevier Ltd · author reuse

Approach

Can a model delineate small kidney stones more accurately by narrowing the search before segmentation?

Two classifiers first identify kidney-containing slices and then stone-containing slices. The kidney stone segmentation network, KSSNet, processes the selected images and reuses features learned from kidney tissue segmentation.

  1. Identify kidney-containing CT slices
  2. Select slices containing stones
  3. Trace the stone boundary with KSSNet
Axial abdominal CT slice with a class activation overlay.
A class-activation map from the study repository. The overlay highlights regions that influenced the slice classifier.View full-size figureImage source

Reported results

Slice selection improved the same KSSNet model. Its stone Dice score rose from 90.13% without scope reduction to 93.10% with kidney-slice selection only and 97.44% with both stages. The slice classifier reached an overall F1 score of 94.45%.

Dice with full scope reduction
97.44%
Dice without scope reduction
90.13%
Subjects in the source CT cohort
259

My contribution

I designed the successive scope-reduction and feature-transfer approach and led model development, experiments, analysis, and manuscript preparation.

Study scope

The five folds separate subjects. The scans came from a single hospital, so the reported scores describe that cohort and protocol.