Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CNN-FL for Biotechnology Industry Empowered by Internet-of-BioNano Things and Digital Twins

Mohammad, Jamshidi, Dinh Thai Hoang, Diep N. Nguyen

TL;DR

This paper addresses the challenge of creating accurate digital twins (DTs) for micro- and nano-scale biological systems by integrating the Internet of Bio-Nano Things (IoBNT) with Convolutional Neural Networks (CNN) and Federated Learning (FL). The proposed CNN-FL-IoBNT framework collects image-based biological data from nano-scale devices, trains a robust global model without sharing raw data, and yields precise DTs of microorganisms, such as bacteria. Simulation on the DIBaS dataset demonstrates that CNN-FL closely matches centralized CNN performance while preserving privacy and reducing data transmission, highlighting improvements in reliability, safety, and energy efficiency. The framework has broad implications for bioprocess design, real-time monitoring, and personalized medicine, potentially transforming biotech and clinical workflows through secure, scalable digital twins. Future work outlines open issues in BNT fabrication, IoBNT communication, bio-cyber interfaces, and multi-scale DT modeling to further close the gap between nano-scale sensing and industrial DT deployment.

Abstract

Digital twins (DTs) are revolutionizing the biotechnology industry by enabling sophisticated digital representations of biological assets, microorganisms, drug development processes, and digital health applications. However, digital twinning at micro and nano scales, particularly in modeling complex entities like bacteria, presents significant challenges in terms of requiring advanced Internet of Things (IoT) infrastructure and computing approaches to achieve enhanced accuracy and scalability. In this work, we propose a novel framework that integrates the Internet of Bio-Nano Things (IoBNT) with advanced machine learning techniques, specifically convolutional neural networks (CNN) and federated learning (FL), to effectively tackle the identified challenges. Within our framework, IoBNT devices are deployed to gather image-based biological data across various physical environments, leveraging the strong capabilities of CNNs for robust machine vision and pattern recognition. Subsequently, FL is utilized to aggregate insights from these disparate data sources, creating a refined global model that continually enhances accuracy and predictive reliability, which is crucial for the effective deployment of DTs in biotechnology. The primary contribution is the development of a novel framework that synergistically combines CNN and FL, augmented by the capabilities of the IoBNT. This novel approach is specifically tailored to enhancing DTs in the biotechnology industry. The results showcase enhancements in the reliability and safety of microorganism DTs, while preserving their accuracy. Furthermore, the proposed framework excels in energy efficiency and security, offering a user-friendly and adaptable solution. This broadens its applicability across diverse sectors, including biotechnology and pharmaceutical industries, as well as clinical and hospital settings.

CNN-FL for Biotechnology Industry Empowered by Internet-of-BioNano Things and Digital Twins

TL;DR

This paper addresses the challenge of creating accurate digital twins (DTs) for micro- and nano-scale biological systems by integrating the Internet of Bio-Nano Things (IoBNT) with Convolutional Neural Networks (CNN) and Federated Learning (FL). The proposed CNN-FL-IoBNT framework collects image-based biological data from nano-scale devices, trains a robust global model without sharing raw data, and yields precise DTs of microorganisms, such as bacteria. Simulation on the DIBaS dataset demonstrates that CNN-FL closely matches centralized CNN performance while preserving privacy and reducing data transmission, highlighting improvements in reliability, safety, and energy efficiency. The framework has broad implications for bioprocess design, real-time monitoring, and personalized medicine, potentially transforming biotech and clinical workflows through secure, scalable digital twins. Future work outlines open issues in BNT fabrication, IoBNT communication, bio-cyber interfaces, and multi-scale DT modeling to further close the gap between nano-scale sensing and industrial DT deployment.

Abstract

Digital twins (DTs) are revolutionizing the biotechnology industry by enabling sophisticated digital representations of biological assets, microorganisms, drug development processes, and digital health applications. However, digital twinning at micro and nano scales, particularly in modeling complex entities like bacteria, presents significant challenges in terms of requiring advanced Internet of Things (IoT) infrastructure and computing approaches to achieve enhanced accuracy and scalability. In this work, we propose a novel framework that integrates the Internet of Bio-Nano Things (IoBNT) with advanced machine learning techniques, specifically convolutional neural networks (CNN) and federated learning (FL), to effectively tackle the identified challenges. Within our framework, IoBNT devices are deployed to gather image-based biological data across various physical environments, leveraging the strong capabilities of CNNs for robust machine vision and pattern recognition. Subsequently, FL is utilized to aggregate insights from these disparate data sources, creating a refined global model that continually enhances accuracy and predictive reliability, which is crucial for the effective deployment of DTs in biotechnology. The primary contribution is the development of a novel framework that synergistically combines CNN and FL, augmented by the capabilities of the IoBNT. This novel approach is specifically tailored to enhancing DTs in the biotechnology industry. The results showcase enhancements in the reliability and safety of microorganism DTs, while preserving their accuracy. Furthermore, the proposed framework excels in energy efficiency and security, offering a user-friendly and adaptable solution. This broadens its applicability across diverse sectors, including biotechnology and pharmaceutical industries, as well as clinical and hospital settings.
Paper Structure (22 sections, 4 figures)

This paper contains 22 sections, 4 figures.

Table of Contents

  1. Introduction
  2. DTs for Biotechnology Industry
  3. Designing and Testing Novel Bioprocesses and Bioproducts
  4. Optimizing Performance, Quality, and Safety
  5. Monitoring and Predicting Behaviors and Outcomes
  6. Holistic View of the Bioprocess Life Cycle
  7. IoBNT and its Potentials for DT-empowered Biotechnology Industry Applications
  8. Internet of Bio-Nano Things (IoBNT): An Overview
  9. Potential Applications of IoBNT for DT-Enabled Biotechnology Industry
  10. Real-Time Data Acquisition and Integration
  11. Improved Model Accuracy and Predictive Analytics
  12. Personalized Medicine and Healthcare
  13. FNN-CNN for Digital Twin integrated with IoBNT for Biology Applications
  14. Proposed System Model
  15. Proposed Solution
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Illustrating how integrating IoBNT can significantly improve applications in the biotechnology industry through DT technology.
  • Figure 2: The convergence of IoBNT and FL for advanced DTs in biotechnology: micro-level data analysis with ML-driven techniques.
  • Figure 3: The framework for creating bacteria DTs, utilizing CNN and FL integrated with IoBNT technology. This framework features a cost-effective approach for microscopic photography, compatible with wireless communication protocols. The system encompasses various clients including clinics and hospitals, contributing to a global model that supports the seamless realization of a sustainable, interactive DT environment.
  • Figure 4: The evaluation and visualization of the proposed approach: (a) The confusion matrix of the conventional CNN trained by the whole of the dataset. (b) The confusion matrix of the proposed method for a third part of data in each area. (c) A detailed physical representation of the bacteria under study, highlighting specific morphological characteristics relevant to the IoBNT application. (d) The DT of the selected group of bacteria indicated by the black circle in (c).