Overview and Objective

Along with the quick evolvement of artificial intelligence (AI), deep/machine learning, and big data in healthcare, medical AI research goes beyond methodological/algorithm development. As of early 2022, the FDA has authorized about 150 medical AI software, and many new research questions are emerging in the practical and applied aspects of medical AI, such as translational study, clinical evaluation, real-world use cases of AI systems, etc. Clinicians are playing an increasingly stronger role in the frontiers of applied AI through collaboration with AI experts, data scientists, informatics officers, and the industry workforce.

Practical applications of medical AI bring in new challenges and opportunities. Today is the best time to strengthen the connections between the arising aspects of AI translation and applications and classic methodological/algorithmic research. The AMAI workshop aims to engage medical AI practitioners and bring more application flavor in clinical, evaluation, human-AI collaboration, new technical strategy, trustfulness, etc., to augment the research and development on the application aspects of medical AI, on top of pure technical research. 

The goal of AMAI is to create a forum to bring together researchers, clinicians, domain experts, AI practitioners, industry, and students to investigate and discuss various aspects related to applications of medical AI. AMAI will 1) introduce emerging medical AI research topics and novel methodology towards applications, 2) showcase the evaluation, translation, use case, and success of AI in healthcare, 3) develop multi-disciplinary collaborations and academic-industry partnerships, and 4) provide educational, networking, and career opportunities for attendees including clinicians, scientists, trainees, and students. The first AMAI workshop will take place in September 2022 in Singapore as a Satellite Event of MICCAI 2022.

AMAI will be composed of invited talks, presentations (oral and poster) of contribution papers/abstracts, and expert panel discussions. AMAI will feature invited paired talks from collaborative computational AI experts and physicians. Each year, AMAI will dedicate a special session centered around discussing the local medical AI efforts of the country where AMAI is held (i.e., Singapore, in 2022) to engage interactions and promote education in medical AI and data sciences in the local communities. 

Call for Submissions

As medical AI is a multi-disciplinary subject, AMAI calls for submissions from multiple aspects of research topics, such as, but not limited to, those listed below. AMAI is agnostic to medical data modalities and encourages submissions using imaging or non-imaging data. 

• Clinical and translational AI/ML applications focusing on addressing specific diseases or medical contexts

• Medical AI/ML evaluation studies in various settings (including simulation and real-world clinical environments)

• Novel AI/ML strategies, approaches, methodologies, tools, and software towards practical utilities 

• Human reader/observer study, human-AI interaction/collaboration, synergy of AI/ML and human/medical intelligence

• Success stories, use cases, challenges, opportunities, lessons learned, or prospects in medical AI/ML applications at work

• Strategies, techniques, evaluation, measurement, reviews, or opinions on clinical useability, explainability, trustworthiness, safety, acceptance, limitations, regulations, or bias of medical AI/ML

• Medical AI/ML use from perspectives of researchers, clinicians, health providers, patients, and the public society

• Multi-disciplinary AI/ML collaboration work among data scientists, clinicians, and other domain experts 

Submissions may be in two tracks:

Full papers: Submissions must be new work. All submissions will be reviewed by at least two experts with experience of relevance. Accepted papers will be assigned as oral or poster presentations primarily based on merit. Each paper will allow a maximum of 8 pages (including texts, figures, and tables) for scientific content and up to 2 additional pages for references. Submissions should be formatted in Lecture Notes in Computer Science (LNCS) style (Please use Springer LaTeX or Word templates) and anonymized for double-blind review. Supplemental materials are not allowed. For accepted papers, the corresponding/senior authors will need to complete and sign a Consent-to-Publish form on behalf of all the authors. For papers invited for publication in the partnering journals, the authors will be asked to convert the accepted papers to align with the format requirements of the journal.

Abstracts: Submissions may be new work or recently published/accepted papers (including posted preprints). All submissions will be reviewed in terms of scientific merit, relevance to the workshop, and significance to the field. Accepted abstracts will be assigned primarily as poster presentations. Submissions will allow a maximum of 1 page (including figures/tables, if any), following specified formats: 0.5-inch margins in all directions; Arial font, 11 points; single column and single line spacing; with the following sections in order: Indication of new or published work (where and when); Title; Authors (including emails of corresponding authors); Purpose; Materials and Methods; Results; Conclusions; Medical/clinical significance; Relevance to AMAI; References (if any). 

All submissions should be submitted via the CMT system: https://cmt3.research.microsoft.com/AMAI2022

Camera-ready Submission Instructions 

Full papers: Please follow the MICCAI main conference's general guidelines (if applicable) for camera-ready submissions: https://conferences.miccai.org/2022/en/CAMERA-READY-GUIDELINES.html. Paper length: a maximum of 8.5 pages (including texts, figures, and tables) for scientific content and up to 2 additional pages for references (this is consistent with the MICCAI main conference). Supplemental materials are not allowed. 

Note: Within the last section (i.e., Discussion or Conclusion) of your paper, it is required to include a separate paragraph at the end of the section to briefly describe the prospect of application of your work. Please refer to below format (note that the phrase of "Prospect of application" needs to be bold font):

Prospect of application: Use maximum 60 words to describe the prospect and envisioned contexts, scenarios, or circumstances on the potential application/deployment of your work. 

The License to Publish form needs to be signed by the corresponding/senior author on behalf of all the authors. The corresponding author The corresponding author signing the copyright form should match the corresponding author marked on the paper. Conference Name (i.e., AMAI 2022) and the Volume Editors' Names are already entered for you at the first page, where you need to fill in the title and names of the all authors and corresponding authors of your paper. This form must be signed in wet-ink. Digital signature is not acceptable. Please scan your signed form and save it as a PDF file (file name format: AMAI2022_License-to-Publish_PaperID.pdf) and upload it to the CMT system. 

The corresponding author must be available to carry out a final proof check of the typeset paper before publishing in the LNCS proceedings. He or she will be given a 72-hour time-slot to do so. The corresponding author should be clearly marked as such in the header of the paper. He or she is also the one who signs the license-to-publish form on behalf of all of the authors. Please note that the corresponding author cannot be changed after the camera ready submission deadline. We encourage the inclusion of all of the authors’ email addresses in the header, but at the very least, the email address of the corresponding author should be present.

For papers invited for publication in the partnering journals, the authors will be asked to convert the accepted papers to align with the format requirements of the journal. This will be a separate process and details will be communicated individually with the authors.

Abstracts: Final version of the accepted abstracts should be formatted strictly following the AMAI Abstract Template. The maximum length of an abstract is 1 page (including everything). Both a Word source document (.docx) and a corresponding PDF document (.pdf) are required as final files to submit to the CMT system. File names should be formatted as: AMAI2022_Abstract_submissionID.docx and  AMAI2022_Abstract_submissionID.pdf. Supplemental materials are not allowed. No copyright form will need to be signed. The accepted abstracts will be made publicly accessible on this website.

Please use the same CMT system link (see above) to submit the final papers or abstracts.


Presentation Instructions

1) The workshop venue/room will be Virgo 1. The workshop will start at 8am on Sep 18  (SGT).

2) Authors of accepted submissions are assumed to be attending the workshop in person, except for those who had been communicated individually for virtual attendance due to critical needs.

3) Your presentation mode (oral or poster) has been notified to you in the acceptance emails.

4) Oral presentations will be a total of 10 minutes per paper (including at least 1 minute at the end of your presentation for questions). There are no specific formatting requirements on the presentation slides. On Sep 18 morning, presenters are asked to show up a bit earlier in the workshop room to copy the presentation files to the computer before the start of the workshop. For virtual presenters, you will present through sharing your screen in the Zoom room.

5) For poster presentations, please make your physical posters referring to the MICCAI main conference formats, size, and requirements (see this link: https://conferences.miccai.org/2022/en/INFORMATION-FOR-PRESENTERS.html). Each poster will be assigned a poster board number. Posters will be staying up throughout the entire workshop hours (possibly longer before they need to be removed) on Sep 18. There will be a dedicated poster session in the workshop agenda for authors to present their posters, but authors can and are encouraged to present/interact anytime while the posters are in display. In order to accommodate virtual participation, for the full papers accepted as poster presentation and for all accepted abstracts, we ask the authors to upload their posters (in PDF format; maximum file size 3 MB) to the CMT system (use the Camera Ready submission link to upload a separate single file of your poster in PDF format) by Sep 12 (11:59pm PST) so that we will provide a link (to be shared to you at the beginning of the workshop) to all registered participants to access the online posters during the workshop time period - this way, all authors/participants will have a chance to view these posters in a more flexible manner and interact with presenters through the Chat function of Zoom or offline. The final version of all the accepted abstracts will be posted in this website too prior to the workshop.

6) We will have a dedicated Zoom room accessible via Pathable for the entirety of the scheduled time of our workshop. The entire workshop will be recorded and registered participants may access the video recording afterwards.

Publishing Plan and Awards

Accepted full papers will be published by Springer Nature as a part of the MICCAI Satellite Events joint LNCS proceedings. Depending on the topic, select full papers will be invited for publication in the journal of Radiology: Artificial Intelligence. Among the accepted full papers, AMAI will give awards to the best papers with certification and cash (provided by sponsors).

Accepted abstracts will not be formally published by publishers (neither in the MICCAI Satellite Events joint LNCS proceedings nor in the partnering journals); they will be made publicly accessible on this website. 

Important Dates

AMAI2022 will be held as a Satellite Event during the conference of MICCAI2022 in Singapore.

Workshop time: September 18, 8:00AM to 11:30AM (SGT time), 2022  

Submissions open: April 25, 2022

Submissions close: 11:59pm, Pacific Time, June 15, 2022

Submissions close: 11:59pm, Pacific Time, June 20, 2022

Notification of acceptance: July 16, 2022

Notification of acceptance: July 20, 2022

Camera ready submission due: 11:59pm, Pacific Time, July 30, 2022

LNCS full paper camera ready submission due: 11:59pm, Pacific Time, Aug. 1, 2022

Abstract final version submission due: 11:59pm, Pacific Time, Aug. 3, 2022

Program Committee 

(In alphabetical order)

• Dooman Arefan, PhD, University of Pittsburgh, USA

• Bettina Baessler, MD, University Hospital Wuerzburg, Germany

• Douglas Hartman, MD, University of Pittsburgh Medical Center, USA

• Yu Jing Jan Heng, PhD, Harvard/Beth Israel Deaconess Medical Center, USA

• Mireia Crispin Ortuzar, PhD, University of Cambridge and Cancer Research, UK

• Chang Min Park, MD, PhD, Seoul National University Hospital, Seoul Korea

• Adam Perer, PhD, Carnegie Mellon University, USA

• Nicholas Petrick, PhD, U.S. Food and Drug Administration, USA

• Zhiyong (Sean) Xie, PhD, Pfizer Inc., USA

• Yudong Zhang, MD, PhD, The First Affiliated Hospital, Nanjing Medical University, China 

Organizers and Sponsors

• • Shandong Wu, PhD, Associate Professor, University of Pittsburgh, USA; Email: wus3@upmc.edu 

  • Behrouz Shabestari, PhD, Director, National Technology Centers Program, National Institute of Biomedical Imaging and Bioengineering (NIBIB), USA; Email: behrouz.shabestari@nih.gov 

  • Lei Xing, PhD, Jacob Haimson & Sarah S. Donaldson Professor, Stanford University, USA; Email: lei@stanford.edu 

  • Niketa Chotai, MD, Radiologist, RadLink Imaging Centre and National University of Singapore, Singapore; Email: niketachotai@gmail.com  

Sponsors

Pittsburgh Center for Artificial Intelligence Innovation in Medical Imaging

Agenda

Time: Sep 18 2022; Sunday, 8am – 11:30am (SGT time)

Onsite venue: Room Virgo 1 (go down stairs/elevator to L1, turn left, follow signs, and walk straightforward)

Virtual attendance is available for registered users through the Pathable system (https://miccai2022.pathable.eu/; it is open to workshop organizers and will be open to regular users on Sep 13), where a dedicated Zoom room will be assigned to our workshop. Note that the link for the Zoom will appear in Pathable until the moment shortly before the workshop starts at the designated time.

On-site posters: Poser boards have been put in the workshop room with labels indicating the event name. The boards are numbered and double sided and some of the boards will be placed in front of the room. It will be easy for authors to find the poster boards. Authors are required to hung up your posters any time but before the start of the workshop. The MICCAI organizers are not assigning a poster a specific number of the board, so authors can find any empty boards in the room to hung up their posters.

• 8:00-8:05am: Introductory remarks: AMAI Organizers

• 8:05-8:25: Keynote talk: Shandong Wu, PhD, Associate Professor, University of Pittsburgh

“Application of Medical AI: Emerging Research Topics”

• 8:25-9:30: Full Paper Oral Presentation Session I: 6 papers (10-11 mins /paper; including presentation and Q&A)

1)     Automated Assessment of Renal Calculi in Serial Computed Tomography Scans

Pritam Mukherjee, Sungwon Lee, Perry J. Pickhardt, and Ronald M. Summers

2)     Deep Learning Meets Computational Fluid Dynamics to Assess CAD in CCTA

Filip Malawski, Jarosław Gośliński, Mikołaj Stryja, Katarzyna Jesionek, Marcin Kostur, Karol Miszalski-Jamka, and Jakub Nalepa      

3)     Machine learning for dynamically predicting the onset of renal replacement therapy in chronic kidney disease patients using claims data

Daniel Lopez-Martinez, Christina Chen, and Ming-Jun Chen

4)     Efficient Neighbor Context-aware Breast Cancer Classification in Digital Breast Tomosynthesis using Transformers (Invited to publish in Radiology: AI)

Weonsuk Lee, Hyeonsoo Lee, HyunJae Lee, Eunkyung Park, Hyeonseob Nam, Thijs Kooi

5)     Increasing the Accessibility of Peripheral Artery Disease Screening with Deep Learning [Virtual presentation]  

Adrit Rao and Oliver Aalami 

6)     A Deep Learning-based Interactive Medical Image Segmentation Framework [Virtual presentation]

Ivan Mikhailov, Benoit Chauveau, Nicolas Bourdel, and Adrien Bartoli

• 9:30-10:10 Poster presentation (8 full papers and 14 abstracts) and coffee break (MICCAI official coffee break time: 9:30-10am) 

Onsite in person presentations of posters (poser board number will be shared to authors once available). 

Virtual participants will interact through Zoom Chat box while viewing digital posters accessible on AMAI’s website.

• 10:10-10:25: Invited talk: Matthew Pease, MD, Memorial Sloan Kettering Cancer Center

“Machine Learning for Traumatic Brain Injury Prognostication and A Surgeon’s Perspective on Working with Data Scientists”

• 10:25-11:00: Full Paper Oral Presentation Session II: 3 papers (10-11 mins /paper; including presentation and Q&A)

7)     OOOE: Only-One-Object-Exists Assumption to Find Very Small Objects in Chest Radiographs

Gunhee Nam, Taesoo Kim, Sanghyup Lee, and Thijs Kooi

8)     CADIA: a Success Story in Breast Cancer Diagnosis with Digital Pathology and AI Image Analysis

María Jesús García-González, Rodrigo Cilla Ugarte, Blanca Zufiria Gerbolés, Kristin May Rebescher, Esther Albertin Marco, Iván Lalaguna, Javier García Navas, Maria Blanca Cimadevila Alvarez, Iván Macía Oliver, Karen López-Linares Román, and Valery Naranjo Ornedo

9)     Wavelet Guided 3D Deep Model to improve Dental Microfracture Detection

Pranjal Sahu, Jared Vicory, Matt McCormick, Asma Khan, Hassem Geha, and Beatriz Paniagua

• 11:00-11:15: Special invited talk: Bhanu Prakash K.N. PhD, Bioinformatics Institute, Agency for Science, Technology and Research (A*STAR), Singapore

     “An Overview of Local Medical AI Efforts at Singapore”

• 11:15-11:25: Additional Q&A and discussion for/from all attendees

• 11:25-11:30: Award announcement (Best Paper, Best Student Paper, and Best Abstract) and closing remarks

• 11:30am: Adjourn

Accepted Papers and Abstracts

Full paper track (oral presentation): 9 papers

See those listed in Agenda 

Full paper track (poster presentation): 8 papers [link to digital posters]

#8 Deep Neural Network Pruning for Nuclei Instance Segmentation in Hematoxylin & Eosin-Stained Histological Images

Amirreza Mahbod, Rahim Entezari, Isabella Ellinger, and Olga Saukh

#12 Was that so hard? Estimating human classification difficulty

Morten Rieger Hannemose, Josefine Vilsbøll Sundgaard, Niels Kvorning Ternov, Rasmus R. Paulsen, and Anders Nymark Christensen

#14 ECG-ATK-GAN: Robustness against Adversarial Attacks on ECGs using Conditional Generative Adversarial Networks

Khondker Fariha Hossain, Sharif Amit Kamran, Alireza Tavakkoli, and Xingjun Ma

#20 Uncertainty-Aware Geographic Atrophy Progression Prediction From Fundus Autofluorescence

Qi Yang, Neha Anegondi, Verena Steffen, Simon S. Gao, Julia Cluceru, Christina Rabe, Jian Dai, and Daniela Ferrara

#27 Prediction of mandibular ORN incidence from 3D radiation dose distribution maps using deep learning      

Laia Humbert-Vidan, Vinod Patel, Robin Andlauer, Andrew P King, and Teresa Guerrero Urbano

#29 Spatial Feature Conservation Networks (SFCNs) for Dilated Convolutions to Improve Breast Cancer Segmentation from DCE-MRI

Hyunseok Seo, Seohee So, Sojin Yun, Seokjun Lee, and Jiseong Barg

#33 Analysis of potential biases on mammography datasets for deep learning model development

Blanca Zufiria, Karen López-Linares, María J. García\, Kristin M. Rebescher, Iván Lalaguna, Esther Albertín, Maria B. Cimadevila, Javier Garcia, Maria J. Ledesma-Carbayo, and Iván Macía

#56 The impact of using voxel-level segmentation metrics on evaluating multifocal prostate cancer localization

Wen Yan, Qianye Yang, Tom Syer, Zhe Min, Shonit Punwani, Mark Emberton, Dean Barratt, Bernard Chiu, and Yipeng Hu

Abstract track (poster presentation): 14 abstracts [link to abstracts]

#4 Discovering distinctive elements of medical image datasets for high-performance exploration

Md Tauhidul Islam, Lei Xing

#5 Effect of testosterone therapy on quantitative breast lobular atrophy in transmasculine individuals

Gabrielle M Baker, Yaileen D. Guzman-Arocho, Vanessa C. Bret-Mounet, Mitko Veta, Suzanne C. Wetstein, Brittany M Charlton, Sarah S. Jackson, Gerburg M Wulf, and Yujing J Heng

#13 Classifying surgical skill level using visual metrics and a gradient boosting classification model

Somayeh B. Shafiei, Saeed Shadpour, James L. Mohler, Gary Smith, Mehdi Seilanian Toussi, Zhe Jing

#15 Fully automated morphometric assessment of Hirschsprung's disease with whole-slide image using deep learning

Changi Kim, Jaemoon Koh, Dayoung Ko, Daseul Park, Youngbin Ahn, Kyung Chul Moon, Sung Hye Park, Hyun-Young Kim, and Young-Gon Kim

#37 Facial Action Unit Detection on Critically ill ICU Patients

Subhash Nerella, Julie Cupka, Patrick Tighe, Azra Bihorac, Parisa Rashidi

#38 A novel framework for segmentation model evaluation – looking beyond Dice Similarity Coefficient

McCague C, Buddenkotte T, Woitek R, Escudero Sánchez L, Hulse D, Pintican R, P. Piyatissa, Rundo L, Bibane-Schonleib C, Oktem O, Sala E, Crispin Ortuzar M

#40 Benchmarking Prognostic Longitudinal Machine Learning Models of Alzheimer's Disease Using Speech Features

Malikeh Ehghaghi, Jekaterina Novikova, Arindam Sett, Mohsen Hejrati, Jessica Robin, Edmond Teng, Somaye Hashemifar

#43 Prognosis of COPD disease in COPDGene study using HRCT and 3D deep neural networks

Qazaleh Mirsharif, Alexandre Coimbra, Claudia Irionde, Somaye Hashemifar

#48 Prediction of Subsequent Fracture in Patients with Hip Fracture Using 2.5D Deep-Learning with Digitally Reconstructed Radiography from Computational Tomography

Yisak Kim, Jung-Wee Park, MD., Byung Woo Kim, Young-Kyun Lee, MD., Sung Hye Kong, MD., Young-Gon Kim, PhD.

#49 Deep high-resolution representation learning for landmark detection in cephalometric X-ray images

Hong Gi Ahn, Tae-Hoon Yong, Wan Kim, and Byungsun Choi

#51 Fully Automated Quantification System of Punctate Epithelial Erosions for The Dry-eye Severity Using Deep-Learning

Daseul Park, Seonghwan Kim, Jewan Lee, Mee Kum Kim, Chang Ho Yoon, Young-Gon Kim

#55 Machine learning and precision metabolic subtyping of obesity: A multi-center serial study

Ziwei Lin, Yao Liu, Chunjun Sheng, Shen Qu

#57 Patient perspectives on medical AI – A mini-review and what’s next for research

Priyanga Gunarathne, Lisa S. Parker, Margaret Rosenzweig, Hayeon Kim, Shandong Wu

#58 Large-scale histological image dataset with various stain conditions and scanners for the robust machine learning model development

Mieko Ochi, Daisuke Komura, Shumpei Ishikawa

Awards

Best Paper Award

Machine learning for dynamically predicting the onset of renal replacement therapy in chronic kidney disease patients using claims data

Daniel Lopez-Martinez, Christina Chen, and Ming-Jun Chen

Google Research, USA

Best Student Paper Award

Increasing the Accessibility of Peripheral Artery Disease Screening with Deep Learning

Adrit Rao (Palo Alto High School) and Oliver Aalami (Stanford University), USA

Best Student Paper Award - Honorable Mention 

A Deep Learning-based Interactive Medical Image Segmentation Framework

Ivan Mikhailov, Benoit Chauveau, Nicolas Bourdel, and Adrien Bartoli

Université Clermont Auvergne, France

Best Abstract Award

Prediction of Subsequent Fracture in Patients with Hip Fracture Using 2.5D Deep-Learning with Digitally Reconstructed Radiography from Computational Tomography

Yisak Kim, Jung-Wee Park, MD., Byung Woo Kim, Young-Kyun Lee, MD., Sung Hye Kong, MD., Young-Gon Kim, PhD.

Seoul National University, Seoul National University Hospital, Seoul National University Bundang Hospital, South Korea