That variability can affect diagnosis and treatment planning, which is why automated approaches have attracted research attention. The team, led by Kanagaraj Suganya and Sundaravadivelu Sumathi with colleagues at institutions in India, built a customised CNN architecture using the sci-kit-learn machine learning package. Their data came from two sources: 267 histopathology images drawn from the public MITOS-ATYPIA-14 dataset, and 31 real-time images from breast cancer patients collected at Sri Ramakrishna Hospital in Coimbatore, for a total of 298 images.

The dataset was randomly divided so that 80 percent of the images trained the model and the remaining 20 percent was reserved for testing and validation. On that held-out data, the model achieved its reported 96.6 percent accuracy, which the authors describe as significantly outperforming traditional methods and existing automated solutions in terms of accuracy and reliability.

Beyond raw accuracy, the researchers emphasise speed. Manual mitotic counting is time-consuming, and the automated approach, they report, significantly reduces the time required for analysis. They also describe the system as cost-effective, a practical consideration for laboratories weighing whether to adopt digital pathology tools.

The study’s authors argue that this advancement holds promise for enhancing diagnostic precision and treatment outcomes for breast cancer patients, ultimately contributing to improved healthcare and patient prognosis. Breast cancer remains one of the leading causes of cancer-related mortality among women worldwide, so tools that make mitosis detection faster and more consistent have clear clinical appeal. As with any machine learning study in medicine, the reported figures reflect performance on the data used in this research, and the dataset of 298 images is modest by the standards of modern deep learning.

The authors themselves frame the result as demonstrating excellent potential for clinical application in breast cancer diagnosis rather than an established diagnostic standard, and further validation beyond this study’s data would be needed before such a system could inform routine practice.

Subject of Research: Automated detection of mitotic figures in breast cancer histopathology images using a convolutional neural network

Article Title: Robust mitosis detection in breast cancer histopathology images using convolutional neural network

Article References: Suganya, K., Sumathi, S., Karthikesh, K., Bhargavi, S., Sethumadhavan, T., Subashini, P., & Krishnaveni, M. (2026). Robust mitosis detection in breast cancer histopathology images using convolutional neural network. Neural Computing and Applications, 38(20), Article 847. https://doi.org/10.1007/s00521-026-12479-0

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12479-0

Keywords: breast cancer, mitosis detection, histopathology, convolutional neural network, deep learning, digital pathology, MITOS-ATYPIA-14, machine learning, diagnostic accuracy, computational pathology, Robust, mitosis