CountAnywhere: frugal cell counting and viability from smartphone hemocytometer photos

In nearly every laboratory where scientists grow living cells, a routine but critical task happens daily: counting them. Before researchers can study how a drug affects a tumor, how a virus infects tissue, or how stem cells turn into new organs, they must know exactly how many cells they have and how many are still alive. For decades, this has been done with a glass slide called a hemocytometer, which holds a tiny drop of liquid over a grid etched into its surface. A scientist places this slide under a microscope, looks through the eyepiece, and manually tallies the cells in the grid squares. It is a slow, tedious process that relies entirely on the human eye and the patience of the person doing the counting. Because people get tired or distracted, two different scientists might count the same drop of cells and get different numbers. To solve this, many labs now buy expensive automated machines that count cells for them, but these devices cost thousands of dollars and require special disposable slides for every single sample, adding up to significant costs over time.

A team of researchers at the University of Maryland School of Medicine has developed a new way to handle this problem that requires no expensive machine and no special slides. They created a free, browser-based tool called CountAnywhere that turns a standard smartphone into a cell counter. The system works by taking a photograph of the glass slide through a microscope and using the grid already etched on the slide to do the math. Instead of relying on a dedicated instrument, the software uses the grid lines as a ruler to measure the area, checks if the image is sharp enough to be useful, and then automatically identifies and counts every cell in the picture. It also determines whether each cell is alive or dead by checking how dark or light it appears compared to its immediate surroundings. This approach removes the need for costly equipment and disposable materials, making accurate cell counting accessible to any laboratory with a microscope and a phone.

The researchers built this system to work with the ordinary photographs that scientists already take when they look at cells. When a user uploads four pictures of the grid squares from a standard glass slide, the software first finds the grid lines in the image. It uses these lines to correct any tilt or distortion in the photo, ensuring that the area being measured is exactly one square millimeter, regardless of how far the phone was from the microscope or what kind of phone was used. This step is crucial because it turns a simple snapshot into a precise measurement tool. Once the image is corrected, the software scans the area to find the cells. In early tests, a general-purpose computer program designed to find objects in images missed many of the cells, especially when they were crowded together or looked very similar to the background. To fix this, the researchers trained the software specifically on images of five different types of cells, showing it thousands of examples of what these cells look like in a hemocytometer. This training allowed the software to recognize the cells with high accuracy, finding nearly every single one even in crowded conditions.

After the software has identified the cells, it decides which ones are alive and which ones are dead. In this method, scientists often mix the cells with a blue dye that only enters cells with damaged membranes, which are the dead ones. The living cells remain bright and clear, while the dead ones turn dark blue. The software measures the brightness of each cell and compares it to the brightness of the liquid right next to it. If the cell is darker than its surroundings, the program marks it as dead; if it is brighter, it marks it as alive. This rule is simple and does not require complex calculations or adjustments for different types of cells. The researchers tested this method on thousands of cells from five different cell lines and found that it correctly identified live and dead cells with a level of accuracy that matched human experts.

To see how well this new method worked in the real world, the researchers compared it against both manual counting by humans and counting by a commercial automated machine. They analyzed 128 different sections of cell samples. The results showed that the smartphone app agreed with the true count more closely than the human counters did. When people counted the cells by hand, they tended to miss more cells as the number of cells in the sample increased, leading to lower counts than were actually present. The app, however, did not make this mistake; it counted the same proportion of cells whether the sample was sparse or very crowded. In terms of speed, the app was also significantly faster. A full count of a sample took about 50 seconds using the app, whereas manual counting took an average of 122 seconds. The commercial machine was the fastest at about 17 seconds, but it required a dedicated instrument and a disposable slide for every sample, whereas the app used only the reusable glass slide and a phone.

The researchers validated their findings across multiple sessions using different types of cells and different smartphone models, including both Android and Apple devices. They confirmed that the system works reliably as long as the image is in focus and the grid is visible. They also noted that the software can be run on a local computer within a laboratory, meaning it does not even need an internet connection to function, which is important for labs in remote areas or those with strict data security rules. The study concludes that this method democratizes cell counting by removing the financial and technical barriers that have kept accurate, automated counting out of reach for many smaller laboratories. By using the grid that is already present on the glass slide, the system transforms a standard piece of laboratory equipment into a powerful, precise, and affordable tool for measuring life at the cellular level.