AI Achieves Human-Level Accuracy in Early Cancer Detection

AI Achieves Human-Level Accuracy in Early Cancer Detection

A revolutionary development in medical imaging has seen an artificial intelligence (AI) achieve accuracy comparable to human experts in detecting early-stage cancer. This new model surpasses existing computer vision technologies, heralding a significant advancement for clinical applications.

The breakthrough comes from Dipti Deb, a PhD scholar at NIT Rourkela, whose new AI-based system named BreastHistoNet has demonstrated remarkable accuracy in detecting early-stage breast cancer. Her model integrates mammogram, ultrasound and histopathology images, achieving a level of diagnostic precision comparable to seasoned radiologists and indeed surpassed existing computer-vision technologies.

This achievement earned her the Best Poster Award at SocProS 2025 and received international acclaim at the IEEE EMBS Conference in Copenhagen 

Medical imaging has long been a cornerstone in the early detection and diagnosis of various cancers. Traditionally, radiologists and specialists have relied on their expertise to interpret these images. However, the advent of AI has brought transformative potential to this field, enhancing accuracy, efficiency, and early detection capabilities.

The new AI model leverages advanced algorithms and extensive training on large datasets to identify cancerous anomalies that might be imperceptible to the human eye. Its performance in clinical trials has not only matched but sometimes exceeded human-level accuracy, offering a promising tool for early intervention and better patient outcomes.

As medical practitioners strive for improved diagnostic precision, the integration of AI into clinical practices could prove to be a game-changer. With further development and validation, this technology holds the promise of becoming a standard part of medical imaging, providing support and augmenting the expertise of healthcare professionals worldwide.

Why this matters

While radiologists remain central to cancer diagnosis, AI systems like BreastHistoNet are proving capable of identifying subtle patterns that may elude human perception, potentially reducing missed early-stage cancers. By offering automated, high-precision analysis across multiple imaging types, this approach could support earlier intervention and improve patient outcomes.

Key Data and Breakthrough Details

  • AI Model: BreastHistoNet
  • Developer: Dipti Deb, PhD scholar at NIT Rourkela
  • Imaging Integration: Mammogram, Ultrasound, and Histopathology images
  • Diagnostic Accuracy: Achieved 94% detection accuracy, matching or surpassing experienced radiologists in early-stage breast cancer diagnosis
  • Clinical Recognition:
    • Awarded Best Poster at SocProS 2025
    • Presented and acclaimed at the IEEE EMBS Conference in Copenhagen

Notable Study Outcomes and Comparative Data

  • Danish Screening Study:
    • Standalone AI cancer detection sensitivity: 63.7%
    • Combined human + AI reading: 73.9% sensitivity
      [See large-scale multi-country analysis for benchmarking: PubMed/Denmark]
  • German Multi-Site Study:
    • 17.6% increase in detection rate with AI-supported screening (6.7 per 1,000 compared with baseline)
      [See Nature, 2025]
  • Visual Diagnostic Performance:
    • Radboud University:
      • Radiologist area under ROC curve (AUC): 0.93 unaided, 0.97 with AI-assist

Reference Links

Latest Tech and AI Posts