3. Results
3.1. Comparison of effectiveness and timein manual and automatic counting
The biggest disadvantage of manual yeast cell counting with a haemocytometer is that the process is very time consuming and inefficient in batch processing; thus, the first assessment index of the automatic counting procedure is its efficiency. Manual and automatic counting were used simultaneously to analyse the same yeast samples in this study regardless of whether the samples were in simple or complex backgrounds. According to the principle of manual haemocytometer counting, in a complex background, it took an average of at least 1-2 minutes to count cells in one middle square, which was further divided into 16 smaller squares, and the total time for one sample was approximately 5-10 minutes by trained observers. To reduce errors, the counting process was repeated three times, which made the total analysis time for each sample approximately 15-30 minutes, and the total counting time for 12 samples was at least 2 hours. In contrast, when yeast cells were counted automatically, a trained operator took approximately 5-10 minutes to take 36 images (3 images per sample included 400 small squares that needed to be counted), 2-5 minutes to set the parameters, and 1-2 minutes for the automatic PC calculation. Therefore, automatic counting in this study took only 2 to 5 percent of the time required for manual counting, which greatly improved the efficiency of the experiment (Figure 3A). These processing times can be further improved after the first operation for samples in the same backgrounds due to the use of the reproducible “Yeast Counter” macro, which made the average counting time for one sample 2-5 minutes. These results demonstrated the high efficiency of yeast cell counting with ilastik and ImageJ in a complex background.
3.2.Higher accuracy exhibited by the automatic method
Subjective factors can play a role in the yeast cell counting process with a haemocytometer; for example, some cells can be half-in/half-out of the square being counted, and it is difficult to be consistent throughout manual counting. Moreover, for budding yeast cells, the standard of sub-cells with different sizes that can be counted as a single individual varies by different observers, and the same observer would also obtain different statistical results. As the software standard can be unified by the operator, we speculated that the automatic yeast cell counting method would be more accurate than manual counting. Thus, the spotting test was applied in this study. Briefly, 5 out of 12 yeast samples in complex backgrounds or not were taken and diluted to the same concentration and then subjected to 10-fold dilutions. As the number of yeast cells is theoretically consistent at the same dilution, it is obvious that the results of manual counting exhibited a greater deviation between samples (Figure 2A and 2C), while those of automatic counting showed better consistency (Figure 2B and 2D) whether the background was complex or not. We also calculated the CFUs (colony-forming units) of the last spot on YPD media, which were anticipated to be 10 in theory according to the formula for concentration. The results revealed that the average value of the CFUs in automatic counting was 7.8, while the value for manual counting was 1.8 (Figure 3B), demonstrating the higher accuracy of the automatic method.