Individualized Melanoma Prediction Using Self-Adaptive Machine Learning
Description: Melanoma is a serious form of skin cancer and can spread to other organs rapidly. The survival rate for patients with melanoma diagnosed at an early stage is much higher than in advanced-stage melanoma when the cancer has spread to lymph nodes or distant organs. However, despite recent advances in the treatment of melanoma, little progress has been made to facilitate its early detection. This is partially due to an already overburdened healthcare system with insufficient availability of healthcare providers to effectively screen the entire at-risk population. To reduce this burden, our proposal leverages large-scale data from electronic health record databases and recent advances in machine learning to identify and triage individuals at highest risk for developing melanoma. We propose to accomplish this goal through the following steps.
1. We aim to develop a machine learning based prediction tool which will output a risk likelihood of a patient for developing melanoma using the patient’s medical records as input.
2. The predicted risk score will be dynamically updated. Namely, whenever a patient’s medical records are updated, the risk score will be updated accordingly. This will enable healthcare providers to identify and enroll high-risk patients in melanoma screening programs in a timely manner.
3. Since the data fed into the prediction tool is collected from routine office visits, an individual can be evaluated for melanoma risk without performing a specific assessment, and the tool can be scaled with minimal cost.
4. With a large multi-institutional data registry of 6,049,538 patients, we have sufficient data for model development and validation. Recent advances in machine learning, especially data representation mining, will enable automatically feature extraction and model selection.
We expect that our prediction tool can be applied clinically to guide individualized screening strategies for melanoma and facilitate detection of melanoma at early-stages.
Funder: Melanoma Research Alliance
Amount: $35,000
Funding type: award
Start date: 2022-07-01
End date: 2023-06-30
Investigators: Guihong Wan, Christopher Decelles, Yevgeniy Semenov, Sharon Germana, Nga Nguyen
Institution: Massachusetts General Hospital (Mass General)
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Funded outputs count: 1
Funder award ID: 988187
OpenAlex ID: G7759397133