We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Computer Science > Computer Vision and Pattern Recognition

Title: Towards Automated Melanoma Screening: Proper Computer Vision & Reliable Results

Abstract: In this paper we survey, analyze and criticize current art on automated melanoma screening, reimplementing a baseline technique, and proposing two novel ones. Melanoma, although highly curable when detected early, ends as one of the most dangerous types of cancer, due to delayed diagnosis and treatment. Its incidence is soaring, much faster than the number of trained professionals able to diagnose it. Automated screening appears as an alternative to make the most of those professionals, focusing their time on the patients at risk while safely discharging the other patients. However, the potential of automated melanoma diagnosis is currently unfulfilled, due to the emphasis of current literature on outdated computer vision models. Even more problematic is the irreproducibility of current art. We show how streamlined pipelines based upon current Computer Vision outperform conventional models - a model based on an advanced bags of words reaches an AUC of 84.6%, and a model based on deep neural networks reaches 89.3%, while the baseline (a classical bag of words) stays at 81.2%. We also initiate a dialog to improve reproducibility in our community
Comments: Minor corrections on State of the Art and Conclusion
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:1604.04024 [cs.CV]
  (or arXiv:1604.04024v3 [cs.CV] for this version)

Submission history

From: Eduardo Valle [view email]
[v1] Thu, 14 Apr 2016 03:26:28 GMT (5782kb,D)
[v2] Wed, 20 Apr 2016 02:45:27 GMT (3401kb,D)
[v3] Fri, 6 May 2016 21:00:22 GMT (3373kb,D)

Link back to: arXiv, form interface, contact.