EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

An Agentic AI Pipeline for Real-Time E. coli Detection and Water Safety Decision-Making Using Photonic Crystal Fiber Biosensors

B. Naveen Kumar

Journal of Intelligent Decision Making and Information Science · 03.10.2026 · 3 (15s) · 2881-2897

DOI: 10.17762/jidmis.v3.4843 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Intelligent Decision Making and Information Science, online erschienen am 03.10.2026. Verfasst von B. Naveen Kumar. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Microbially contaminated drinking water causes an estimated 500,000 diarrheal deaths each year, and Escherichia coli is the principal indicator of fecal contamination. Reference methods are slow or laboratory-bound: culture requires 24–48 h, polymerase chain reaction (PCR) requires 2–4 h with thermal cycling, and lateral flow assays (LFAs) return neither quantitative nor regulation-aligned results. This study presents a deterministic five-agent pipeline that converts the 12-parameter optical–electrical output of a photonic crystal fiber (PCF) biosensor into an explained, threshold-referenced water-safety decision. A Sensor Agent validates inputs against COMSOL Multiphysics simulation bounds; an ML Agent executes three Random Forest models for binary detection, eight-class concentration classification and log-concentration regression; a Reasoning Agent generates template-based explanations from feature importances; a Decision Agent maps the concentration estimate to tiers derived from WHO and US EPA criteria; and an Alert Agent compiles a structured report. On a held-out split of 600 COMSOL-grounded synthetic samples, binary detection reached 95.17% accuracy (AUC-ROC = 0.993), eight-class classification 80.33% (weighted F1 = 0.811) and regression R^2=0.918. Against a refractive-index (RI) threshold baseline evaluated on the same split (67.67%), this is a gain of 27.50 percentage points. Mean end-to-end latency was 67.49 ms per sample on a single CPU thread (n = 200 runs). The explanation and decision layers add regulatory traceability without altering classification accuracy. Because all data derive from simulation with modeled noise, the results establish computational feasibility; validation with fabricated sensors and real water samples is required before field deployment.

Zitieren

APA

Kumar, B. N. (2026). An Agentic AI Pipeline for Real-Time E. coli Detection and Water Safety Decision-Making Using Photonic Crystal Fiber Biosensors. Journal of Intelligent Decision Making and Information Science, 3(15s), 2881-2897. https://doi.org/10.17762/jidmis.v3.4843

Vancouver

Kumar BN. An Agentic AI Pipeline for Real-Time E. coli Detection and Water Safety Decision-Making Using Photonic Crystal Fiber Biosensors. Journal of Intelligent Decision Making and Information Science. 2026;3(15s):2881-2897. doi:10.17762/jidmis.v3.4843

BibTeX
@article{Kumar2026,
  author = {Kumar, B. Naveen},
  title = {An Agentic AI Pipeline for Real-Time E. coli Detection and Water Safety Decision-Making Using Photonic Crystal Fiber Biosensors},
  journal = {Journal of Intelligent Decision Making and Information Science},
  year = {2026},
  volume = {3},
  number = {15s},
  pages = {2881--2897},
  doi = {10.17762/jidmis.v3.4843}
}
Kostenlos starten →