Title: Investigating the utilisation of Field-Programmable Gate Arrays (FPGA) Camera-Based approach for Neonatal Jaundice detection

Introduction:

Neonatal jaundice is a common condition that affects up to 60% of newborns worldwide, causing yellowing of the skin and eyes due to high levels of bilirubin in the blood. Although usually harmless, in severe cases, it can cause brain damage and even death. Early detection and treatment are crucial to prevent these complications. However, the current methods of diagnosing neonatal jaundice (by visual inspection or by measuring bilirubin levels in the blood) are either invasive, expensive, or inaccurate, limiting their availability and effectiveness in many geographical regions, especially in third world nations. There is a need for a non-invasive, low-cost but still accurate method for jaundice detection and one promising approach is the use of Field-Programmable Gate Arrays (FPGA) and a camera.

This research aims to investigate the utilisation (FPGA) Camera-Based approach for neonatal jaundice detection. FPGAs are reconfigurable hardware chips that can be programmed and reprogrammed to implement customised logic circuits. They can perform parallel processing and image processing tasks efficiently and flexibly. By using a camera attached to an FPGA board, the skin colour of a newborn can be captured and analysed to estimate the bilirubin level and potentially diagnose jaundice.

FPGAs are a viable solution for the computation of neonatal jaundice detection because they offer several advantages over other hardware devices. First, FPGAs are reconfigurable, meaning that they can be programmed and modified according to the specific needs and requirements of the application. They can perform parallel processing and image processing tasks at high speed and low power consumption. This enables fast and accurate analysis of the skin colour images captured by the camera.

The end goal is to open up grounds for more advanced implementation of this research project that will eventually improve the quality and accessibility of Neonatal jaundice detection, especially in low-resource settings where conventional methods are not available or affordable.

Methodology & Timeline:

To carry out this project, I plan to design and implement an algorithm to imitate a Bilirubinometer for the FPGA hardware, which will read and analyse in real time the image input from the integrated camera.

Firstly, a simple FPGA design will be implemented that uses the corresponding camera IPs inside Libero, a software tool for FPGA design. The goal is to be able to read the raw data from the cameras in real time from the FPGA. 

Following the camera’s integration with the FPGA, I will begin by designing the parallel algorithms for implementing the first FPGA design of the Bilirubinometer. Due to limited time constraints, the Bilirubinometer, a device typically used to measure bilirubin levels in the blood, will be adapted to a more focused scope, resembling a streamlined skin color analysis system. A Bilirubinometer traditionally measures the concentration of bilirubin by analyzing the skin color, as bilirubin causes a yellowish hue in the skin when levels are elevated. By shining a beam of light on the skin of the newborn, the light is absorped and the intensity of the light reflected back is calculated.  Implementing this in my research, I will utilize the integrated camera and FPGA setup to capture images of the newborn's skin. These images will then be processed using algorithms developed specifically for this purpose (streamlined skin colour analysis system). For example, a skin colorimeter algorithm will use a color space conversion technique to extract the hue component from the RGB image and compare it with a predefined threshold to determine the presence of jaundice.

This proposed approach has its limitations. Traditional Bilirubinometers rely on specialized equipment and precise calibration procedures for accurate measurements, which may be challenging to replicate with a simplified skin color analysis system. Additionally, accurately measuring bilirubin levels through image analysis alone can be difficult due to factors like varying lighting conditions and skin tones. The accuracy of the skin colorimeter algorithm depends on the quality of the captured image and the effectiveness of the color space conversion technique, which may not always provide reliable results. To improve the accuracy of this method, I will have to carry out constant testing, refining algorithms, and undertaking validation studies to mitigate for the constraints of this approach.

The Verilog design of the parallel algorithm will be set up to work with the FPGA board so I will synthesise the developed algorithm for the target FPGA device then integrate the synthesised algorithm with the FPGA hardware. This will ensure proper functionality as the goal at this time is accuracy of the algorithm. Tests will be carried out to validate the performance of the FPGA-implemented algorithm, assessing its accuracy and efficiency in jaundice detection tasks.

The FPGA Camera-Based system will also be tested and its accuracy evaluated based on real time image capturing. The goal here is to achieve real time operation and functionality of the algorithm on the FPGA Camera-Based system.

By analysing the data and comparing the results using relevant metrics, I will record the findings and implications of the research and highlight the strengths and limitations of the proposed FPGA Camera-Based system. Finally, after drawing concussions and reflecting, recommendations for future work, improvements and extensions of the FPGA Camera-Based system will be made.

Intended Outcomes:

• Analyse and conclude on the feasibility of using FPGA Camera-Based system for neonatal jaundice detection.
• Contribute my research findings for future research and advancements in the field of FPGA detection systems.
• Start the process of creating on open source hardware for lower barrier to entry for future advancements in the space.