Ultrastructural Morphometry of Mitochondria: Comparison Between Conventional Operator-Dependent and Artificial...

Area of Science:

• Cell Biology
• Microscopy
• Artificial Intelligence

Background:

• Morphometric analysis provides objective data for scientific research.
• Artificial intelligence (AI) and machine learning (ML) enable automated morphometry.
• Adapting AI for transmission electron microscopy (TEM) morphometry is challenging due to complex ultrastructural details.

Purpose of the Study:

• To evaluate the accuracy of ML-based mitochondrial morphometry against manual analysis using TEM images.
• To compare the quantitative results of automated and manual methods for mitochondrial cristae surface area ratio (C/A ratio).

Main Methods:

• Comparison of ML-automated and manual morphometric analysis on TEM micrographs (×50,000 magnification).
• Analysis of cultured cells (n=26) with varying energy metabolism.
• Calculation of the mitochondrial cristae surface area ratio (C/A ratio) as a measure of mitochondrial function.

Main Results:

• No statistically significant correlation was found between ML-automated and manual morphometry methods.
• ML-derived data showed greater variability (scatter) than manual measurements.
• Current ML protocols cannot reliably distinguish subtle ultrastructural details crucial for accurate TEM morphometry.

Conclusions:

• ML-based automated morphometry is currently not accurate enough for TEM-based mitochondrial analysis.
• Expert human interpretation remains essential for recognizing and analyzing complex cellular ultrastructure in TEM images.
• Further development of ML algorithms is needed to improve accuracy in specialized microscopy applications.