We use cookies to understand how you use our site and to improve your experience. This includes personalizing content and advertising. To learn more, click here. By continuing to use our site, you accept our use of cookies. Cookie Policy.

Features Partner Sites Information LinkXpress hp
Sign In
Advertise with Us
GLOBETECH PUBLISHING LLC

Download Mobile App




Ultra-Lightweight AI Model Runs Without GPU to Break Barriers in Lung Cancer Diagnosis

By MedImaging International staff writers
Posted on 18 Jun 2025
Image: Schematic representation of the ultra-lightweight AI model architecture and training process (Photo courtesy of Kenji Suzuki/Institute of Science Tokyo)
Image: Schematic representation of the ultra-lightweight AI model architecture and training process (Photo courtesy of Kenji Suzuki/Institute of Science Tokyo)

Artificial intelligence (AI) models typically demand enormous datasets and expensive GPU servers, creating a significant barrier to wider adoption, especially in resource-limited settings. Now, researchers have developed an ultra-lightweight deep learning model that can assist in lung cancer diagnosis using only a standard laptop and a minimal dataset, offering a highly accessible alternative to traditional AI-driven diagnostic systems.

This revolutionary AI model was developed by a research team from the Institute of Science Tokyo (Tokyo, Japan) using a unique deep learning approach based on massive-training artificial neural network (MTANN). This enabled them to build a powerful diagnostic model without the high computational cost and data volume typically associated with deep learning. Unlike conventional large-scale AI models that rely on training with thousands of medical images, this model uniquely learns from individual pixel data extracted from computed tomography (CT) scans. This approach significantly reduced the required dataset size to just 68 cases. The entire training process was completed in just 8 minutes and 20 seconds on a standard laptop, showcasing the model’s extraordinary efficiency.

Despite being trained on a small dataset and run on limited hardware, the model demonstrated exceptional diagnostic performance. It achieved an area under the curve (AUC) value of 0.92, far outperforming state-of-the-art large-scale AI models like Vision Transformer and 3D ResNet, which only achieved AUC values of 0.53 and 0.59, respectively. Once trained, the model delivered diagnostic predictions at a rapid speed of 47 milliseconds per case, making it practical for real-time clinical use. This ultra-lightweight AI solution offers substantial practical advantages. It eliminates the need for costly infrastructure and massive datasets, making advanced AI-driven lung cancer diagnostics more accessible to smaller hospitals, rural clinics, and developing regions. Its rapid analysis speed and high accuracy also have the potential to reduce diagnostic delays and support early intervention, which is critical for improving lung cancer outcomes.

“This technology isn’t just about making AI cheaper or faster,” said Kenji Suzuki from the Institute of Science Tokyo, who led the research team that developed the model. “It’s about making powerful diagnostic tools accessible, especially for rare diseases where training data is hard to obtain. Furthermore, it will cut down the power demands for developing and using AI at data centers substantially, and can solve the global power shortage problem we may face due to the rapid growth in AI use.”

Related Links:
Institute of Science Tokyo

Ultrasound Needle Guidance System
SonoSite L25
Adjustable Mobile Barrier
M-458
Computed Tomography System
Aquilion ONE / INSIGHT Edition
Biopsy Software
Affirm® Contrast

Channels

Nuclear Medicine

view channel
Image: Perovskite crystal boules are grown in carefully controlled conditions from the melt (Photo courtesy of Mercouri Kanatzidis/Northwestern University)

New Camera Sees Inside Human Body for Enhanced Scanning and Diagnosis

Nuclear medicine scans like single-photon emission computed tomography (SPECT) allow doctors to observe heart function, track blood flow, and detect hidden diseases. However, current detectors are either... Read more

General/Advanced Imaging

view channel
Image: Concept of the photo-thermoresponsive SCNPs (J F Thümmler et al., Commun Chem (2025). DOI: 10.1038/s42004-025-01518-x)

New Ultrasmall, Light-Sensitive Nanoparticles Could Serve as Contrast Agents

Medical imaging technologies face ongoing challenges in capturing accurate, detailed views of internal processes, especially in conditions like cancer, where tracking disease development and treatment... Read more

Imaging IT

view channel
Image: The new Medical Imaging Suite makes healthcare imaging data more accessible, interoperable and useful (Photo courtesy of Google Cloud)

New Google Cloud Medical Imaging Suite Makes Imaging Healthcare Data More Accessible

Medical imaging is a critical tool used to diagnose patients, and there are billions of medical images scanned globally each year. Imaging data accounts for about 90% of all healthcare data1 and, until... Read more
Copyright © 2000-2025 Globetech Media. All rights reserved.