Cutting-Edge AI for Healthcare: Optimized Pneumonia Diagnosis on MAX78000 Microcontroller

¹ Department of Electronics and Communication Engineering, SRM University, Andhra Pradesh, India
² Department of Electronics and Communication Engineering, SRM University, Andhra Pradesh, India
³ Department of Electronics and Communication Engineering, SRM University, Andhra Pradesh, India

^* Corresponding author: rahulgowtham_poola@srmap.edu.in

Abstract

The methodology discussed in the research herein results in the design of a pneumonia diagnosis system that can be executed in real time on hardware edge board. The methodology includes the design and optimization of a neural network classification model for a given pneumonia classifier within the scope of the study. The specifics of the neural network architecture are based upon the designed architecture, which has been modified to fit the computation requirements of the MAX78000 microcontroller. Once trained, the model is exported to file formats supported by the MAX78000 processor, which allows it to be used in the Eclipse programming environment. The final point is the deployment of the trained neural network model with a parallel camera and display which eliminates the need for external software and enables real-time pneumonia diagnosis. This integration emphasizes the flexibility and robustness of the system, and it demonstrates how a low-power system as the MAX78000 allows rethinking the approaches towards the medical diagnostics. After the system is deployed, it is able to deliver accurate predictions where it achieved 97% normal class prediction and 88% pneumonic class prediction. These results reaffirm the system’s efficacy in real-time classification tasks, highlighting the diagnostic potential.