Predict lesion segmentation masks from dermoscopic images. Approximately 2600 images for training, 100 for validation, 1000 for test. $2,500 prize for top performing algorithm.
Skin is the largest organ of the human body, and is the first area of assessment performed by any clinical staff when a patient is seen, as it delivers numerous insights into a patient’s underlying health. For example, cardiac function, liver function, immune function, and physical injuries can be assessed by examining the skin. In addition, dermatologic complaints are also among the most prevalent in primary care . Images of the skin are the most easily captured form of medical image in healthcare, and the domain shares qualities to other standard computer vision datasets, serving as a natural bridge between standard computer vision tasks and medical applications.
This workshop will serve as a venue to facilitate advancements and knowledge dissemination in the field of skin image analysis, as well as host a melanoma detection challenge, raising awareness and interest for these socially valuable tasks. Invited speakers include (confirmed) major influencers in computer vision and skin imaging, top ranked participants of the hosted challenge, and accepted papers on skin image analysis.
Topics of manuscript submissions may include:
- Deep Learning for Dermatology
- Computer Vision for Dermatology
- Dermoscopic Image Analysis
- Consumer-Grade Photograph Analysis
- Video Analysis
- Video or Image Processing
- Total-Body Photography
- Confocal Microscopy Analysis
- Optical Coherence Tomography (OCT) Analysis
- Skin Histopathological Image Analysis
- Other Imaging Modalities Related to Skin
- Interpretable or Explainable AI Related to Skin Image Analysis
The workshop will give out a "Best Paper" award (with $2,500 prize), as well as awards and prizes to winners of the associated challenge.
 Lowell BA, et al. “Dermatology in primary care: Prevalence and patient disposition” In: Journal of the American Academy of Dermatology (JAAD), vol. 45, no. 2. 2001.
This workshop will feature several prominent names in the field of skin image analysis, including:
|Olivier Gevaert||Assistant Professor of Medicine and of Biomedical Data Science, Stanford Center for Biomedical Informatics Research (BMIR), Stanford University.|
|Josep Malvehy||Dermatology Department, Hospital Clínic of Barcelona, University of Barcelona, Spain|
|Rafael Garcia||Computer Vision and Robotics Institute, University of Girona, Spain|
|Laura Ferris||Associate Professor of Dermatology at University of Pittsburg.|
|Tracy Petrie||Product Manager at OHSU.|
|Gabriela Oana Cula||Research Associate Director at Johnson and Johnson.|
This workshop will feature presentations from both submitted papers and challenge participants.
|April 2nd, 2018:||Challenge Training Data Release|
|NEW: June 30th, 2018:||Workshop Paper Submission Deadline: 11:59 PM EDT|
|July 9th, 2018:||Challenge Validation and Test Data Release|
|July 27th, 2018:||Challenge Submission Deadline|
|Sept 20th, 2018:||Workshop day, Room Albeniz|
Skin cancer is the most common form of cancer, according to the American Cancer Society. While skin cancer is amenable to early detection by direct inspection, visual similarity with benign lesions makes the task difficult. The challenge hosted in conjunction with this workshop will be the newest instance of the largest challenge in the community for automated skin cancer recognition, hosted by the International Skin Imaging Collaboration (ISIC). The goal of the challenge is to develop methods for segmentation, clinical attribute detection, and disease classification in dermoscopic images.
In conjunction with the machine learning challenge, we will conduct a large reader study focused on disease diagnosis, orchestrated by the International Dermoscopy Society (IDS). The study is anticipated to provide human diagnostic accuracy data for hundreds of images from the test set, collected from hundreds of clinician participants worldwide. The data will be made available, stratified by years of clinical experience and geography. This data will serve as a baseline to compare the diagnostic performance of machine learning algorithms to clinical experts of varying skill.
The ISIC 2018 Challenge incorporates three independent sub-tasks:
Classify and localize clinical dermoscopic attribute patterns as binary masks. Categories include "network", "negative network", "streaks", "milia-like cysts", "dots and globules". Approximately 2600 images for training, 100 for validation, 1000 for test. $2,500 prize for top performing algorithm.
Classify disease categories for dermoscopic images. Possible disease states are "Melanoma", "Melanocytic nevus", "Basal cell carcinoma", "Actinic keratosis / Bowen’s disease (intraepithelial carcinoma)", "Benign keratosis (solar lentigo / seborrheic keratosis / lichen planus-like keratosis)", "Dermatofibroma", and "Vascular lesion". Approximately 10,000 images provided for training, 200 for validation, 1500 for test. $2,500 prize for top performing algorithm.
Full information, including data download links, is available at the Challenge Website.
Outline of the workshop schedule. TBD
Workshop and Challenge:
- M. Emre Celebi Ph.D. (University of Central Arkansas, Arkansas, USA)
- Noel C. F. Codella Ph.D. (IBM Research, New York, USA) Twitter
- Kristin Dana Ph.D. (Rutgers University, New Jersey, USA)
- David Gutman M.D. (Emory University, Georgia, USA)
- Allan Halpern M.D. (Memorial Sloan Kettering Cancer Center, New York, USA)
- Brian Helba (Kitware, New York, USA)
- Harald Kittler M.D. (Medical University of Vienna)
- Philipp Tschandl, M.D. Ph.D. (Medical University of Vienna, Vienna, Austria)
- Rogerio Feris Ph.D. (IBM Research, New York, USA)
- Anthony Hoogs Ph.D. (Kitware, New York, USA)
- John R. Smith Ph.D. (IBM Research, New York, USA)