A Deep Learning Framework for Detection and Localization in Molecular Communication

Molecular communication (MC) is a bio-inspired communication paradigm in which information is transmitted through chemical molecules rather than electromagnetic signals. It has emerged as a promising technology for nanoscale biomedical applications, particularly in molecular communication–based targeted drug delivery (TDD) systems, where nanomachines detect abnormalities and release therapeutic agents at specific locations inside the body. Accurate detection and localization of diseased cells are essential in such systems because conventional drug delivery methods often lack precise sensing and targeting capability, leading to inefficient treatment and potential damage to healthy tissues. Despite extensive research in molecular communication, many existing studies focus on signal detection, localization, or communication modelling independently. These approaches often rely on simplified analytical models that struggle to capture the stochastic diffusion behaviour and environmental noise present in biological systems. To address this limitation, this study aims to develop an intelligent framework capable of performing simultaneous abnormality detection and three-dimensional localization in diffusion-based molecular communication environments for targeted drug delivery. A physics-aware simulation environment was constructed in which molecular propagation follows Fick’s law of diffusion with Poisson noise in a three-dimensional spherical medium. Using the generated dataset, a hybrid CNN–BiLSTM deep learning architecture was designed to extract spatial diffusion patterns using convolutional layers and temporal molecular arrival dynamics using bidirectional LSTM layers.Simulation results demonstrate that the proposed framework achieves detection accuracy up to 98.5% with an AUC close to 0.98 and localization error around 0.79–0.816 µm, outperforming trilateration-based and LSTM-only baseline approaches. These findings highlight the effectiveness of hybrid deep learning models for modelling complex diffusion signals and improving detection reliability. Overall, the proposed framework contributes to the advancement of AI-driven targeted drug delivery systems, supporting future developments in nanomedicine, Internet of Bio-Nano Things (IoBNT), and intelligent biomedical communication networks