Institute of Science Tokyo just launched a new AI model that diagnoses lung cancer on a regular laptop. No GPUs, no huge datasets needed. The team led by Professor Kenji Suzuki rolled out this lightweight deep learning model at the RSNA 2024 Annual Meeting.
This AI skips big data. Instead of thousands of images, it learned directly from pixels of just 68 CT scans. Despite tiny training, it scored an area under the curve (AUC) of 0.92 — way better than Vision Transformer and 3D ResNet, which got 0.53 and 0.59.
Training took only 8 minutes and 20 seconds on a standard laptop. Predictions come in at 47 milliseconds per case. This means cancer diagnosis can happen fast, cheap, and anywhere. 
The innovation snagged the Cum Laude Award at RSNA 2024, an honor given to only 1.45% of presentations.
Professor Kenji Suzuki commented on the impact:
"This technology isn’t just about making AI cheaper or faster. 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."
Suzuki invented the MTANN deep learning tech in the early 2000s and holds over 40 patents. He’s a top 2% scientist worldwide and a session chair at the AAAI 2024 conference.
This breakthrough could reshape medical AI by making advanced diagnostics affordable and accessible, especially where data and resources are limited.
Source: Institute of Science Tokyo
