Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

A Two-Stage Augmentation Pipeline Combining Median Filter, SRAD, and Unsharp Masking for Enhanced Kidney Ultrasound Segmentation

Rektorat UDINUS 2025 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2025
Goals
Record
12055438-b28b-4bb1-9099-c00571925317
Published
2026-10-30

Universitas Dian Nuswantoro (UDINUS), through activity participants Setiawan, Agung Nugroho; Rustad, Supriadi; Soeleman, Arief; and Budiman, Fikri, conducted and reported a study on preprocessing for kidney ultrasound image segmentation. The activity addressed speckle noise in kidney ultrasound images. The source describes speckle noise as a multiplicative artifact that degrades visual quality, obscures anatomical boundaries, and hampers the accuracy of deep learning models in segmentation tasks.

The activity proposed and evaluated a preprocessing pipeline used as a data augmentation strategy to improve segmentation precision. The pipeline integrated a Median Filter for general noise reduction, Speckle-Reducing Anisotropic Diffusion (SRAD) to address speckle while preserving edges, and Unsharp Masking for sharpness restoration. The activity used a public kidney ultrasound dataset containing 390 training images and 96 validation images, with annotations for four classes of kidney structures.

Two YOLOv8-seg segmentation models were trained and evaluated. The Baseline Model used the original dataset, while the Augmentation Model used the augmented dataset produced through the proposed preprocessing pipeline. The evaluation showed that the Augmentation Model significantly outperformed the Baseline Model in the mAP50-95 metric. The source states that this metric strictly measures boundary delineation precision. The finding indicates that the proposed preprocessing pipeline enhanced the model's ability to identify the boundaries of kidney structures accurately.

The activity concerns health research focused on kidney ultrasound image analysis and segmentation precision. The source further identifies accurate boundary delineation as particularly important for clinical applications such as quantitative monitoring of chronic kidney disease, where accurate morphological measurements are essential. Through this activity, UDINUS documented a preprocessing and augmentation approach for kidney ultrasound segmentation.

Reference https://www.scopus.com/pages/publications/105037898866?origin=resultslist