Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Novel Approach for Testing Water Safety Using Deep Learning Inference of Microscopic Images of Unincubated Water Samples

Sanjay Srinivasan

TL;DR

A solution exceeding UNICEF's ideal Target Product Profile requirements for presence/absence testing, with an estimated per-test cost of \$0.44, and achieving 93\% accuracy, with precision of 90\% and recall exceeding 94\%.

Abstract

Fecal-contaminated water causes diseases and even death. Current microbial water safety tests require pathogen incubation, taking 24-72 hours and costing \$20-\$50 per test. This paper presents a solution (DeepScope) exceeding UNICEF's ideal Target Product Profile requirements for presence/absence testing, with an estimated per-test cost of \$0.44. By eliminating the need for pathogen incubation, DeepScope reduces testing time by over 98\%. In DeepScope, a dataset of microscope images of bacteria and water samples was assembled. An innovative augmentation technique, generating up to 21 trillion images from a single microscope image, was developed. Four convolutional neural network models were developed using transfer learning and regularization techniques, then evaluated on a field-test dataset comprising 100,000 microscope images of unseen, real-world water samples collected from fourteen different water sources across Sammamish, WA. Precision-recall analysis showed the DeepScope model achieves 93\% accuracy, with precision of 90\% and recall exceeding 94\%. The DeepScope model was deployed on a web server, and mobile applications for Android and iOS were developed, enabling Internet-based or smartphone-based water safety testing, with results obtained in seconds.

A Novel Approach for Testing Water Safety Using Deep Learning Inference of Microscopic Images of Unincubated Water Samples

TL;DR

A solution exceeding UNICEF's ideal Target Product Profile requirements for presence/absence testing, with an estimated per-test cost of \$0.44, and achieving 93\% accuracy, with precision of 90\% and recall exceeding 94\%.

Abstract

Fecal-contaminated water causes diseases and even death. Current microbial water safety tests require pathogen incubation, taking 24-72 hours and costing \50 per test. This paper presents a solution (DeepScope) exceeding UNICEF's ideal Target Product Profile requirements for presence/absence testing, with an estimated per-test cost of \$0.44. By eliminating the need for pathogen incubation, DeepScope reduces testing time by over 98\%. In DeepScope, a dataset of microscope images of bacteria and water samples was assembled. An innovative augmentation technique, generating up to 21 trillion images from a single microscope image, was developed. Four convolutional neural network models were developed using transfer learning and regularization techniques, then evaluated on a field-test dataset comprising 100,000 microscope images of unseen, real-world water samples collected from fourteen different water sources across Sammamish, WA. Precision-recall analysis showed the DeepScope model achieves 93\% accuracy, with precision of 90\% and recall exceeding 94\%. The DeepScope model was deployed on a web server, and mobile applications for Android and iOS were developed, enabling Internet-based or smartphone-based water safety testing, with results obtained in seconds.
Paper Structure (13 sections, 4 equations, 11 figures, 5 tables)

This paper contains 13 sections, 4 equations, 11 figures, 5 tables.

Table of Contents

  1. Introduction and Related Work
  2. Engineering Goals
  3. Methodology: Microscopy of Water Samples
  4. Methodology: Deep Learning Dataset Construction
  5. Foundational Dataset
  6. Augmentations to Dataset
  7. Field Testing Dataset
  8. Deep Learning Model Construction and Results
  9. Model Performance
  10. DeepScope Model Distribution and Results
  11. Model Distribution via Internet
  12. Model Distribution via Smartphone App
  13. Discussion and Conclusions

Figures (11)

  • Figure 1: Image of Rainwater Puddle Underneath Compound Microscope at 1000x Magnification
  • Figure 2: Image of Shower Drain Water Underneath Compound Microscope Stained with Methylene Blue Dye at 1000x Magnification.
  • Figure 3: Illustration of image augmentation. The image is divided into 16 tiles, and random pairs of tiles are selected and their contents swapped.
  • Figure 4: Locations of water bodies in Sammamish where water samples were collected for field testing.
  • Figure 5: Microscope images at 1000x magnification of water sample from Sources 5 (left) and 6 (right). Both sources were found to be unsafe in color change coliform testing.
  • ...and 6 more figures