Estimating the hole surface area of insecticide-treated nets using image analysis, manual hole counting and exact hole measurements

Emmanuel Mbuba, Natalia Mañas-Chavernas, Sarah Moore, Philipo David Ruzige, Dickson Kobe, Rose Philipo, Noela Kisoka, Gianpaolo Pontiggia, Frank Chacky, Charles Dismas Mwalimu, Philippe C. Cattin, Julia Wolleb, Robin Sandkuehler, Amanda RossMalaria 2025 · DOI

Abstract

Images coupled with manual segmentation contain sufficient information to calculate hole surface area. This lays the groundwork for incorporating automatic hole detection, and then assessing whether this method will offer a fast and objective method for routine assessment of physical integrity of ITNs. While the WHO method underestimated the hole surface area, it remains useful in classifying nets as either serviceable or too torn because the cut-off is specific to this method.

Introduction: The Challenge of Monitoring ITN Integrity

Insecticide-treated nets (ITNs) are a cornerstone of malaria prevention. However, their long-term efficacy depends heavily on their physical condition. Currently, the World Health Organization (WHO) recommends a manual counting method to monitor net "health." This involves field teams manually counting holes and categorizing them into four size classes based on common objects (e.g., "smaller than a thumb" or "larger than a head").

Despite being the gold standard, this approach suffers from significant subjectivity, geometric inaccuracies—as it assumes all holes are circular—and high labor intensity. This study investigates whether digital image analysis can provide a more accurate, objective, and scalable alternative for estimating hole surface area.

Methodology

The researchers evaluated three measurement techniques using 75 field-aged ITNs from Bagamoyo, Tanzania, which had been in use for 36 months.

1. The WHO Method

Field teams manually counted and categorized holes by size. The total area was estimated using the midpoint of each size category, assuming circular geometry.

2. Image Analysis via Manual Segmentation

Nets were mounted on a specialized frame for photography. High-resolution images were captured of each side against a high-contrast background.

Assembly of the metal frame used for standardized net photography.

The setup utilized a black-clothed frame with a white grid to ensure the net was flat and the holes were clearly visible.

A net side mounted on the black-clothed frame with a white grid, ready for image capture.

Researchers then used a custom Graphical User Interface (GUI) to manually trace (segment) the holes in these photographs to calculate their area.

3. Ground Truth Reference

To establish a definitive baseline, calibrated close-up shots were taken of every hole larger than 0.5 cm. ArUcO markers—square fiducial markers with known physical dimensions—were placed next to each hole. This allowed for precise perspective correction and exact area calculation in square centimeters.

Key Insights and Results

The study revealed several critical findings regarding the accuracy of current monitoring practices:

  • Failure of the Circular Assumption: The WHO method's reliance on circular geometry introduces systematic bias, as real-world net damage is often irregular or elongated. Image analysis captures the true geometry, avoiding this error.
  • Systematic Underestimation: Contrary to some previous findings, the WHO method was found to significantly underestimate damage, particularly for nets with holes larger than 10 cm. This is likely due to the lack of an upper size limit for "Category 4" holes.
  • High Information Density: A key "proof-of-principle" was that a single photograph taken from 250 cm away contains enough pixel data to match the accuracy of high-precision close-up shots. This suggests that the primary challenge for future automation is the development of robust segmentation algorithms rather than image resolution.
  • Statistical Performance: Image analysis showed substantial agreement with the ground truth across a wide range of damage (10 cm² to 10,000 cm²). While it showed a slight, constant underestimation (mean ratio of 0.70), it remained far more consistent than the WHO method, which showed increasing bias as damage worsened.

Conclusion and Future Work

Digital image analysis provides a viable path toward objective ITN monitoring. While the segmentation in this study was manual, the results confirm that the necessary data is present in whole-net photography.

Future efforts will focus on automating this process using Deep Learning models such as Mask R-CNN or YOLO. Additionally, researchers must address "hidden damage" at frame corners and test the performance of these algorithms in less controlled, real-world field conditions, such as dim lighting or wrinkled fabric in household settings.

Summary updated

Abstract

Images coupled with manual segmentation contain sufficient information to calculate hole surface area. This lays the groundwork for incorporating automatic hole detection, and then assessing whether this method will offer a fast and objective method for routine assessment of physical integrity of ITNs. While the WHO method underestimated the hole surface area, it remains useful in classifying nets as either serviceable or too torn because the cut-off is specific to this method.
BibTeX
@inproceedings{mbuba2025estimating,
  author = {Emmanuel Mbuba and Natalia Mañas-Chavernas and Sarah Moore and Philipo David Ruzige and Dickson Kobe and Sarah Moore and Rose Philipo and Noela Kisoka and Gianpaolo Pontiggia and Frank Chacky and Charles Dismas Mwalimu and Philippe C. Cattin and Julia Wolleb and Robin Sandkuehler and Amanda Ross},
  title = {{Estimating the hole surface area of insecticide-treated nets using image analysis, manual hole counting and exact hole measurements}},
  booktitle = {Malaria},
  year = {2025},
  doi = {10.1186/s12936-025-05324-7},
}
Figure