AI Liver Tumor Segmentation: High Accuracy Model
- Tokyo—A team from the Institute of Science Tokyo, led by Professor Kenji Suzuki and Yuqiao Yang, has developed a new AI model, MHP-Net, that accurately identifies liver tumors...
- The multi-scale Hessian-enhanced patch-based neural network (MHP-Net) overcomes the challenge of big data requirements in medical AI. The model segments medical images into small 3D patches, focusing on...
- The result is a high-resolution tumor segmentation map that accurately delineates liver tumors from contrast-enhanced CT scans.
A revolutionary AI model,MHP-Net,accurately identifies liver tumors from CT scans,even with limited training data,a major breakthrough in medical AI,according to new research. This groundbreaking technology, developed by teh Institute of Science Tokyo, outperforms existing systems in tumor segmentation, achieving remarkable results. MHP-Net’s rapid training and real-time inference capabilities make it ideal for resource-constrained clinical settings. By using a novel patch-based approach, MHP-Net overcomes the need for voluminous datasets, perhaps democratizing AI in healthcare. This advancement will enable scalable, cost-effective AI deployment. Stay informed with more breakthroughs like this one, brought to you by News Directory 3. Discover what’s next for AI in medicine.
AI Model Achieves High Accuracy for Liver Tumor Segmentation
Tokyo—A team from the Institute of Science Tokyo, led by Professor Kenji Suzuki and Yuqiao Yang, has developed a new AI model, MHP-Net, that accurately identifies liver tumors from computed tomography (CT) scans, even when trained with limited data. Their findings were published in IEEE Access.
The multi-scale Hessian-enhanced patch-based neural network (MHP-Net) overcomes the challenge of big data requirements in medical AI. The model segments medical images into small 3D patches, focusing on specific areas. It then pairs each patch with an enhanced version using Hessian filtering, which highlights spherical objects like tumors.
The result is a high-resolution tumor segmentation map that accurately delineates liver tumors from contrast-enhanced CT scans. The model’s performance was evaluated using the Dice similarity score, which measures how well the predicted segmentation matches the ground truth, as steadfast by expert radiologists, on a scale of 0 to 1.
“Despite a limited training set of 7, 14, and 28 tumors, we achieved high performance dice scores of 0.691,0.709, and 0.719, respectively,” Suzuki said. He added that these scores surpass those of established models like U-Net, Res U-Net, and HDense-U-Net.
The model’s lightweight architecture allows for rapid training (under 10 minutes) and real-time inference (approximately 4 seconds per patient), making it suitable for clinical settings with limited resources, according to researchers.
Suzuki noted that this is a starting point for small-data AI, where clinically relevant deep learning models can be built from limited datasets. He believes MHP-Net’s success can inspire similar solutions in other areas of medical imaging, such as detecting rare cancers.
The study highlights the potential of small-data AI in medical image analysis.by reducing the data requirements for training, MHP-Net makes AI more accessible, especially in under-resourced hospitals and clinics. The researchers plan to explore broader applications of small-data AI models to enable scalable, cost-effective, and versatile AI deployment in healthcare worldwide.
More information: Yuqiao Yang et al, Patch-Based Deep-learning Model With Limited Training Dataset for Liver Tumor Segmentation in Contrast-Enhanced Hepatic Computed Tomography, IEEE Access (2025).DOI: 10.1109/ACCESS.2025.3570728