Schematic illustration of the ultra-lightweight synthetic intelligence mannequin structure and coaching course of primarily based on a massive-training synthetic neural community (MTANN). Credit score: Kenji Suzuki from Institute of Science Tokyo, Japan
Think about diagnosing most cancers not with a supercomputer however on an unusual laptop computer as an alternative. Appears like science fiction? Because of a revolutionary synthetic intelligence (AI) mannequin developed by Professor Kenji Suzuki and his analysis group from Institute of Science Tokyo (Science Tokyo), this far-fetched situation is now a actuality.
Unveiled on the Radiological Society of North America (RSNA) 2024 Annual Assembly, the group launched an ultra-lightweight deep studying mannequin that assists with lung most cancers analysis with out counting on expensive graphics processing unit (GPU) servers or huge datasets. Developed utilizing a singular deep studying method primarily based on massive-training synthetic neural community (MTANN), the mannequin was skilled and examined on nothing greater than an ordinary laptop computer laptop, reaching what as soon as required complete knowledge facilities.
AI, skilled by deep studying fashions, has gained important consideration lately, resulting in improvements in a number of fields of analysis. It has additionally been reported that if a deep studying mannequin is skilled on a considerable amount of knowledge, comparable to 1,000,000 photos, it may purchase a efficiency that may surpass that of typical applied sciences and even people.
The place most fashions depend on large knowledge, the AI mannequin developed by Suzuki’s group is exclusive—in contrast to typical large-scale AI fashions, it doesn’t require complete medical picture units. As a substitute, it learns straight from particular person pixels extracted from computed tomography (CT) scan photos. This technique considerably diminished the variety of required instances from 1000’s to only 68!
Regardless of being skilled solely on a small set of knowledge, the mannequin outperformed state-of-the-art large-scale AI methods, comparable to Imaginative and prescient Transformer and 3D ResNet, attaining a discrimination efficiency akin to an space underneath the curve (AUC) worth of 0.92 (towards AUC values of 0.53 and 0.59 for the normal state-of-the-art (SOTA) fashions, respectively).
As soon as skilled, with the complete coaching course of solely taking 8 minutes and 20 seconds on an ordinary laptop computer, it may generate diagnostic predictions at an unprecedented price of 47 milliseconds per case.
Regardless of being skilled on a considerably lower-computational setup (MacBook Air with M1 chip), the 3D Large-Coaching Synthetic Neural Community (MTANN) achieves superior efficiency (space underneath the curve (AUC) = 0.92), sooner inference, and drastically diminished coaching time and parameter depend in comparison with that of 3D ResNet. Credit score: Kenji Suzuki from Institute of Science Tokyo, Japan
“This technology isn’t just about making AI cheaper or faster,” says Suzuki. “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.”
In recognition of its significance, the group’s analysis was conferred the coveted Cum Laude Award at RSNA 2024, an honor acquired by just one.45% of the 1,312 shows. Whereas this innovation is bound to have a transformative impression on most cancers analysis, it stands as a testomony to Suzuki’s deep information and unwavering dedication.
With profound experience within the subject of biomedical AI, Suzuki was the primary to invent the MTANN expertise (used within the present analysis) within the early 2000s. It was one of many earliest deep studying fashions that he had developed and improved on. In his 25 years of analysis expertise, Suzuki has made important contributions to his subject, with greater than 400 publications and over 40 patents, most of which have been licensed and commercialized.
Past this, his current achievements embody serving as a session chair on the thirty ninth Annual AAAI Convention on Synthetic Intelligence. He has acquired two of RSNA’s highest distinctions for his analysis in 2024. Furthermore, he’s acknowledged among the many prime 2% scientists worldwide.
Suzuki continues to steer groundbreaking analysis on the intersection of AI and medical imaging, actively fostering interdisciplinary collaboration that pushes the boundaries of what AI can obtain in scientific follow. His group’s work on compact, high-performance diagnostic fashions exemplifies how progressive pondering—mixed with sensible implementation—can bridge gaps between engineering and medication.
With a dynamic analysis atmosphere and a powerful community of collaborators, Suzuki will not be solely advancing the sector of biomedical AI but additionally serving to form the subsequent technology of translational medical applied sciences.
Supplied by
Institute of Science Tokyo
Quotation:
Most cancers analysis in your laptop computer? New synthetic intelligence mannequin makes it attainable (2025, June 5)
retrieved 6 June 2025
from https://medicalxpress.com/information/2025-06-cancer-diagnosis-laptop-artificial-intelligence.html
This doc is topic to copyright. Aside from any truthful dealing for the aim of personal research or analysis, no
half could also be reproduced with out the written permission. The content material is offered for data functions solely.
