【セミナー】兵庫県立大学・国際研究セミナー
Essam Rashed
Image-MLの皆様、
兵庫県立大学のラシドと申します。
情報科学研究科の国際研究セミナーはハイブリッドで開催されます。
■ 2025/09/30 (火) 13:00~14:00
タイトル：Beyond One-Size-Fits-All: Foundational Models for Organ-Centric Medical Imaging
ゲストスピーカー：Kayhan Batmanghelich (Boston University, USA)
場所: 神戸情報科学キャンパス・大講義室（7階720号）
アクセスマップ：https://www.u-hyogo.ac.jp/about/access/
登録（必須）：https://shorturl.at/GhozE
■ Abstract: The rapid advancement of artificial intelligence has spurred a growing shift toward foundational models, including in applied fields like medical imaging. These models promise to streamline the development process by replacing multiple task-specific models with a single, versatile framework trained on large multimodal data. While this concept is compelling and early results are encouraging, our analysis reveals that current foundational models fall short in addressing the unique complexities of medical imaging. In this talk, I propose a middle-ground solution: organ-specific foundational models tailored to domains such as lung and breast imaging. Drawing from our recent works, Mamo-CLIP and MedSyn, I will highlight both the potential and the limitations of this approach. By addressing key challenges, including data scarcity, annotation burden, and anatomical variability. I will discuss practical strategies for building effective domain-specific foundational models. The talk will conclude with a forward-looking perspective on opportunities to advance foundational model development in medical imaging.
■ Biography: Kayhan Batmanghelich, Ph.D. is an Assistant Professor in the Department of Electrical and Computer Engineering at Boston University. His research focuses on the intersection of artificial intelligence and healthcare, with an emphasis on medical imaging, explainable AI, and multimodal learning. He develops domain-specific foundational models that integrate radiological imaging, clinical data, and molecular information to support diagnosis, prognosis, and therapeutic decision-making. Dr. Batmanghelich has led multiple research projects supported by the NIH, NSF, and industry sponsors, and collaborates closely with clinicians to translate machine learning innovations into clinical workflows. He is a recipient of the Google Faculty Research Award and a Junior Faculty Fellow at the Hariri Institute for Computing.
=========================================
詳細と登録フォームについては、WPをご覧ください。
https://u-hyogo.info/joint-research/seminar/
=========================================
ご参加をお待ちしております。
よろしくお願いします。
-------------------------------------
ラシド イサム
兵庫県立大学大学院情報科学研究科
〒650-0047 神戸市中央区港島南町7-1-28
電話: 078-303-1924 (内線 610)
E-mail: ras...@gsis.u-hyogo.ac.jp
URL: https://u-hyogo.info/research/faculty/erashed/
-------------------------------------
Essam Rashed
From: Essam Rashed <ras...@gsis.u-hyogo.ac.jp>
Date: Sat, Aug 9, 2025 at 11:58 AM
Subject: 【セミナー】兵庫県立大学・国際研究セミナー
To: <m...@image-ml.org>
Image-MLの皆様、
兵庫県立大学のラシドと申します。
情報科学研究科の国際研究セミナーはハイブリッドで開催されます。
■ 2025/09/30 (火) 13:00~14:00
タイトル：Beyond One-Size-Fits-All: Foundational Models for Organ-Centric Medical Imaging
ゲストスピーカー：Kayhan Batmanghelich (Boston University, USA)
場所: 神戸情報科学キャンパス・大講義室（7階720号）
アクセスマップ：https://www.u-hyogo.ac.jp/about/access/
登録（必須）：https://shorturl.at/GhozE
■ Abstract: The rapid advancement of artificial intelligence has spurred a growing shift toward foundational models, including in applied fields like medical imaging. These models promise to streamline the development process by replacing multiple task-specific models with a single, versatile framework trained on large multimodal data. While this concept is compelling and early results are encouraging, our analysis reveals that current foundational models fall short in addressing the unique complexities of medical imaging. In this talk, I propose a middle-ground solution: organ-specific foundational models tailored to domains such as lung and breast imaging. Drawing from our recent works, Mamo-CLIP and MedSyn, I will highlight both the potential and the limitations of this approach. By addressing key challenges, including data scarcity, annotation burden, and anatomical variability. I will discuss practical strategies for building effective domain-specific foundational models. The talk will conclude with a forward-looking perspective on opportunities to advance foundational model development in medical imaging.
■ Biography: Kayhan Batmanghelich, Ph.D. is an Assistant Professor in the Department of Electrical and Computer Engineering at Boston University. His research focuses on the intersection of artificial intelligence and healthcare, with an emphasis on medical imaging, explainable AI, and multimodal learning. He develops domain-specific foundational models that integrate radiological imaging, clinical data, and molecular information to support diagnosis, prognosis, and therapeutic decision-making. Dr. Batmanghelich has led multiple research projects supported by the NIH, NSF, and industry sponsors, and collaborates closely with clinicians to translate machine learning innovations into clinical workflows. He is a recipient of the Google Faculty Research Award and a Junior Faculty Fellow at the Hariri Institute for Computing.
=========================================
詳細と登録フォームについては、WPをご覧ください。
https://u-hyogo-gsis.org/joint-research/seminar/
=========================================
ご参加をお待ちしております。
よろしくお願いします。