Experiment Data & Analysis

Overview

This repository contains raw data, code and analysis scripts related to experiments performed in the ‘A combined microfluidic deep learning approach for lung cancer cell high throughput screening toward automatic cancer screening applications’. The data, code, and documentation provided here to facilitate reproducible research and enable further exploration and analysis of the experimental results.

Repository Contents

Analysis Code:

Languages: MATLAB 2020a or later with Deep Learning Toolbox

Description: This repository contains MATLAB scripts for data preprocessing, deep learning-based classification, and visualization of lung cancer cell images. The scripts train convolutional neural networks (CNNs) to classify six lung cell lines, including normal and five cancer subtypes.

Documentation: 

File: LungCancer_CellLine_Code.zip

Description: This file provides exemplary code and sample images used for the machine learning approach.

File: Supplementary information and instructions.pdf

Description: This file provides an instruction and a description of the individual steps from raw data to image analysis.

File: Original Image data and Metadata Example - pc9.zip

Description: This .zip container provides an example of raw data in a native .vsi file format with folders containing the .ets file, with metadata documentation of the imaging parameters for a microfluidic channel imaged with the IX83 microscope.

File: Data augmentation documentation.docx (and Data augmentation documentation.pdf)

Description: This document provides descriptions of how data augmentation was performed.

File: Raw data.zip

Description: This file contains image raw data.

File:  GrayCellData.rar

Description: This file contains image data converted to grayscale images.

File: CellData_Full.rar

Description: This file contains RGB image data.

Microfluidic cultivation protocol prior to imaging: 

Cell Lines: The lung normal cell and non-small lung cancer cells (PC-9, SK-LU-1, H-1975, A-427, and A-549)

Plate Format: Plasma-bonded and coated microfluidics chip platform fabricated with silicon sheets and sterile object glass slides.

Surface Coating

Prior to cell seeding, the surface of the polydimethylsiloxane (PDMS) microfluidic chip was treated with collagen to enhance cell adhesion. A 0.1% (w/v) collagen solution was prepared using Type I collagen (derived from rat tail) dissolved in a 0.02 M acetic acid buffer. The PDMS surfaces were incubated with the collagen solution for 2 hours at room temperature to allow for proper coating. Following this, the chips were rinsed with phosphate-buffered saline (PBS) to remove any unbound collagen. Collagen, being a key extracellular matrix component, provides a conducive environment for cell attachment and proliferation. This surface modification was crucial for ensuring that the cells would adhere effectively to the microfluidic architecture, promoting optimal growth conditions. The collagen coating facilitated stronger cell-matrix interactions, thereby improving the overall experimental reliability and enabling accurate analysis of cell behavior in the microfluidic system.

Seeding Density

In this study, various cell types (lung normal cells and non-small cell lung cancer cells: PC-9, SK-LU-1, H-1975, A-427, and A-549) were cultured within a microfluidic chip designed with a total length of 75 mm and a width of 25 mm, featuring three separate chambers, each with a diameter of 900 μm. The seeding density was calculated to be approximately 5,000 cells/mL. Given the chamber dimensions, this density was optimized to ensure that the cells could achieve ~70% confluency within a reasonable timeframe while maintaining their viability and functionality. The initial seeding in a 25 cm² culture flask allowed for efficient expansion and preparation of the cells prior to their transfer to the microfluidic environment (the cell culture medium was DMEM or RPMI supplemented with 10% FBS and 1% PS).

Cultivation Duration

After trypsin treatment of cells cultured in a flask, the cells were allowed to adhere to the microfluidic chip for a duration of 48-72 hours post-injection. This incubation period was essential for the cells to establish stable adhesion to the collagen-coated surfaces, enabling them to regain their morphology and functionality. It ensured that the cellular environment within the microfluidic chambers mimicked in vivo conditions, allowing for proper cell spreading and growth.

Medium Composition

The medium utilized for cell cultivation consisted of DMEM (Dulbecco's Modified Eagle Medium) or RPMI-1640, supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (PS), tailored to the specific cell types used. This composition was chosen to provide the necessary nutrients, growth factors, and antibiotics to support cell proliferation and prevent contamination. DMEM and RPMI are known to support a wide range of mammalian cell types, thereby enhancing the versatility of the experimental setup. The medium was pre-warmed to 37°C before use, and the cells were maintained in a humidified incubator at 37°C with 5% CO₂ during cultivation.

Imaging Setup

The imaging data was acquired using an automated IX83 microscope (Olympus, Japan), featuring a Merzhäuser motorized stage, a Hamamatsu ORCA-Flash4.0 camera, and a Lumencolor Spectra X fluorescent light source. This setup ensures high-resolution fluorescence imaging with precise stage control and sensitive image capture. Data was recorded automatically after adjustment of the z-axis using a multi-region area of interest on each microfluidic channel with the focus map function (medium density setting) with cellSens Dimension software (Version 2.1-2.3, Olympus). The DAPI staining of the blue fluorescence channel was used to facilitate large-area adjustment of the focus map prior to automated imaging. The green fluorescence channel representing the phalloidin staining of f-actin was used as a single channel exported images for the deep learning procedure outlined in the paper.

Setup and Installation

1. Extract the Raw Data:

Unzip the Raw data.zip file into your working directory.

2. Environment Setup:

Read the documentation Supplementary information and instructions.pdf and the readme.txt in the code for more details on the setup.

3. Running the Analysis:

Open the file Supplementary information and instructions.pdf for a detailed description.

Usage Instructions

Data Exploration: The analysis scripts include functions for exploratory data analysis (EDA). You can modify these scripts to investigate specific experimental conditions.

Reproducibility

Follow the code comments and documentation to replicate the analyses. Ensure that the environment and dependencies are correctly configured as described in the setup section.

Licensing

This repository is licensed as follows: Code is accessible under BSD 2-Clause "Simplified" license and data under a Creative Commons Attribution 4.0 International license.

Acknowledgement:

This work was supported by the Iran National Science Foundation (INSF) Grant No. 96006759.

Contact persons:

For data acquisition:

Abdullah Allahverdi, a-allahverdi@modares.ac.ir;

Hadi Hashemzadeh, Hashemzadeh.hadi@gmail.com;

Mario Rothbauer, mario.rothbauer@tuwien.ac.at

For data processing and augmentation:

Seyedehsamaneh Shojaei, s.shojaie@irost.ir, samane.shojaie@gmail.com