Digital Pathology using Raspberry Pi
Written by Bryan Zafra on 06.12.2022
Precise diagnosis is very crucial in managing patient’s disease. It influences what course of treatment is needed by the patient. Incorrect diagnosis leads to incorrect management or treatment which will lead to harm. In the field of medicine, precise diagnosis is expensive. Application of histopathological stain alone ranges from $10 to $37 USD; when you include the expert assessment of the pathologist, the price could go as high $300 USD per hour. [1,2,3] This is costly to patients especially if they are making out-of-pocket payments. Med4PAN project tries to create solution to this problem. With the current advancements in the field of digital health, such as machine/deep learning and internet-of-things, we try to leverage these in cutting the cost of histopathological diagnosis by creating an AI model with a near human expert precision. Our hardware setup consists of:
Microscope
C-mount adaptor
Raspberry Pi with modular camera
LED monitor, keyboard, mouse
The Raspberry Pi is a small computer system, at most measuring 8.5 X 5.6 cm in dimension, that can fit different hardware modules. For the setup in med4PAN project, a digital camera with 12 megapixels sensors become a modular component of the Raspberry Pi. Also, the LED monitor, keyboard, and mouse are attached to the Raspberry Pi.
The camera module is attached to the C-mount adaptor which is attached to the trinocular port of the microscope. The histopathology slide is mounted to the microscope stage. The objective lens can be changed for the desired magnification level needed. The view from the objective lens will be transmitted through the trinocular lens to the C-mount adapter that will be received by the modular camera. The Raspberry Pi has its own operating system and software that can generate the image of the histopathology slide. This image can be saved in the memory of the Raspberry Pi and can be sent through the internet. Furthermore, more advanced histopathology image analysis can be done such as image classification and object detection using machine/deep learning.
