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

Download:

Current browse context:

q-bio.QM

Change to browse by:

References & Citations

Bookmark

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

Quantitative Biology > Quantitative Methods

Title: An adaptive treatment recommendation and outcome prediction model for metastatic melanoma

Abstract: Melanoma is a type of skin cancer developed from melanocytes. It is one of the most lethal types of cancer, accounting for approximately 75% of skin cancer deaths. Late stage melanoma is very difficult to treat, since the cancer cells are deranged, may be genetically linked and can be unresponsive to therapy. Therefore, determining how to effectively make use of different treatment regimens is of vital importance to survival. In this analysis, we propose an adaptive treatment recommendation system based on a hybrid cluster-classification (CC) structure. Our proposed system consists of two parts,1) distribution based clustering and 2) classification. Our recommendation system can help to identify high-risk melanoma patients and suggest the best approach to treatment, which enables clinicians and patients to make decisions on the basis of real-world data. Our data came from the Canadian Melanoma Research Network (CMRN) database, a pan-Canadian multi-year observational database, which is part of Global Melanoma Registry Network (GMRN). Training/testing sets are generated based on data from different sources, leading to cross cohort analysis tasks. Experimental results show that our proposed system achieves very promising results with an overall accuracy of up to 80%.
Comments: Machine Learning for Health (ML4H) Workshop at NeurIPS 2018 arXiv:1811.07216
Subjects: Quantitative Methods (q-bio.QM); Medical Physics (physics.med-ph)
Cite as: arXiv:1811.09851 [q-bio.QM]
  (or arXiv:1811.09851v1 [q-bio.QM] for this version)

Submission history

From: Xue Teng [view email]
[v1] Sat, 24 Nov 2018 16:24:01 GMT (18kb,D)

Link back to: arXiv, form interface, contact.