Aillis Wins Back-to-back 1st Place at the Renowned AI conference MICCAI 2025, AI Competition

Global recognition for Aillis’s technological excellence in Surgical Visual Question Answering

Aillis, Inc. (Head Office: Chuo-ku Tokyo, CEO: Sho Okiyama), a developer of AI-based medical devices, is pleased to announce that our AI engineer, Quan Huu Cap, has won first place in a competition held as part of the International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2025)—one of the world’s leading international conferences in the field of medical image analysis and computer-assisted medical intervention. The competition was organized by Intuitive Surgical, the developer and manufacturer of the Da Vinci surgical robot, and focused on developing AI technologies that serve as the “eyes” of robotic systems.This year’s challenge required AI models to identify surgical instruments and procedural actions from surgical video footage. Following last year’s victory in a renal pathology image analysis challenge, this marks Aillis’s second consecutive first-place award at a MICCAI Challenge.

MICCAI (Medical Image Computing and Computer-Assisted Intervention) is a top-tier global conference covering interdisciplinary research areas such as medical imaging, machine learning, robotics, and surgical assistance. It serves as a platform for researchers and companies around the world to present the latest advances in technology that support medical innovation.
 During the conference, the annual “MICCAI Challenges” invite participants to compete on specific AI topics such as medical image analysis and AI-assisted diagnosis, evaluating their models’ performances based on scientific accuracy and innovation.

In this year’s competition, Quan developed a novel approach that leveraged Vision-Language Models (VLMs)—a class of generative AI models that can interpret both visual and textual information. Instead of spending extensive time manually labeling vast amounts of surgical video data, he designed sophisticated prompts that enabled the model to achieve high accuracy without relying on human-annotated video captions. This innovative prompt engineering strategy represents a new paradigm in medical video analysis, reducing the dependence on manual data labeling and expanding the possibilities of AI utilization in medical imaging.

Aillis actively supports our AI engineers in participating in international challenges and competitions. Through these initiatives, we foster an environment where exceptional engineers can advance their technical expertise while driving both personal growth and organizational development. Aillis continues to build a structure where technological excellence and business growth evolve in harmony.

■Comment from Quan Huu Cap

Participating in the Surgical Visual Understanding 2025 challenge has been an exciting experience. This competition gave me the chance to dive into a completely new area of surgical video data and VLMs, which I had very little prior experience with.

The task focused on visual question answering from surgical videos, but the very limited captions and noisy data made it challenging to label and analyze, especially as a solo participant. Toward the end of the competition, I came up with a “prompt engineer” strategy with a few effective tricks that allowed me to achieve the highest score without relying on any ground-truth video captions.

I’m truly honored to have won first place in this challenge. Huge thanks to Aillis for the constant support and for always encouraging me to participate in data science competitions. I hope the techniques developed here can contribute to improving video understanding and assist in advancing the field of surgical AI.

■Comment from Atsushi Fukuda, CTO of Aillis 

We are very proud that Quan has won first place in the international machine learning competition organized by MICCAI for the second consecutive year.

This challenge required AI to interpret surgical videos and identify instruments and ongoing procedures—an extremely complex task. Such technology forms a fundamental basis not only for surgical video understanding but also for applications in fields such as industry and education, and it directly connects to our product development efforts.

Aillis will continue to engage in high-impact AI challenges that contribute to society, particularly in healthcare, and will apply the insights gained to our product development in pursuit of creating better medical experiences for all.

アイリス株式会社に所属するAIエンジニアのQuan Huu Cap(以下：Quan)が、医用画像処理ならびにコンピュータ支援による医療介入の分野でトップクラスの国際会議とされるMICCAI2025の主催する、手術の動画から使用されている手術器具や作業状況をAIに推定させるコンペティションで世界1位となりましたことをお知らせします。昨年の腎臓病理画像解析コンペに続く快挙であり、アイリスとしては2年連続でMICCAI Challenge世界一の栄冠を獲得しました。

■ アイリスについて 

「みんなで共創できる、ひらかれた医療をつくる。」をミッションに掲げ、2017年に創業。現役医師である代表・沖山をはじめ、医療従事者、厚生労働省・経済産業省ほかの行政出身者、AI医療領域に特化したデータサイエンティスト、大手医療機器メーカー出身者など多数のプロフェッショナルが揃い、深層学習技術（AI技術）を活用し、医師のもつ匠の技をデジタル化するAI医療機器を開発しています。

【会社概要】

・会社名：アイリス株式会社

・代表取締役：沖山翔

・事業内容：AIを用いた医療機器の開発・製造・販売及び機械学習技術の研究開発

・設立：2017年11月

・本社所在地：〒104-0028 東京都中央区八重洲2-2-1 八重洲セントラルタワー7階

・企業URL：https://aillis.jp/