AI system automates cancer cell detection in cytology, study shows

AI system automates cancer cell detection

Scientists have developed an artificial intelligence system that can autonomously analyze cells on microscope slides for signs of cancer. The research, published in Nature, presents a method that could overcome major weaknesses in the long-established diagnostic procedure of cytology.

Cytology involves lab workers inspecting cells on glass slides to detect early cancer in organs like the lung and bladder. The process is subjective, labour-intensive, and challenged by workforce shortages. The new AI approach, created by Nitta et al., offers a potential path toward automation.

How the new AI cytology works

The system uses a technique called whole-slide edge tomography. It doesn't just take a flat image of a sample; it captures detailed three-dimensional information about cells by scanning them from multiple angles.

This provides a far richer dataset than traditional 2D microscopy. The AI is then trained to identify and classify cells within this 3D space, distinguishing between healthy and potentially cancerous ones with high accuracy.

The key components of the system include:

• A specialized microscope that rapidly captures multi-angle images
• A deep learning algorithm trained on vast datasets of annotated cell samples
• Software that provides a diagnostic readout, flagging suspicious cells for further review

Addressing a critical bottleneck in medicine

The research arrives as pathology labs globally face a growing shortage of skilled cytotechnologists. The manual screening process is slow, and human fatigue can lead to diagnostic errors.

Previous attempts to automate cytology have struggled because flat, 2D images often lack the detail needed for reliable AI analysis. Cells can overlap or be obscured, leading to false negatives. The 3D imaging approach directly tackles this problem.

In their paper, the authors demonstrate that their system performs at a clinical-grade level, matching or exceeding the detection rates of human experts in controlled studies. It can process slides continuously without fatigue.

The future of automated diagnostics

This technology does not aim to replace pathologists but to act as a powerful assistant. It would handle the initial, tedious screening of slides, flagging only the most concerning cases for a human expert's final diagnosis.

This could drastically reduce workload and allow medical professionals to focus on complex interpretations and patient care. The authors suggest it could be particularly valuable for high-volume screening programs, like for cervical or bladder cancer.

Wider adoption would require rigorous clinical trials and regulatory approval. The system must be validated across diverse patient populations and cancer types. If successful, it represents a significant step toward more efficient, accessible, and consistent cancer diagnostics.