A Deep Learning Approach for Prostate Cancer Screening
Principal Investigator
Name
Julio Cesar Gali
Institution
Pontifícia Universidade Católica de São Paulo
Position Title
Professor of Surgery
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-1440
Initial CDAS Request Approval
Jan 11, 2024
Title
A Deep Learning Approach for Prostate Cancer Screening
Summary
This study will explore the application of a deep neural network (DNN) in prostate cancer screening. Central to the project will be a DNN model trained on patient data, encompassing age, PSA, free PSA levels, and their ratios, to predict the probability of prostate cancer. The model's architecture is planned to feature multiple dense layers with ReLU activation, concluding with a sigmoid output layer for binary classification.
To address the prevalent issue of class imbalance in medical datasets, the project will employ resampling techniques. This strategy aims to ensure a balanced representation of cancerous and non-cancerous cases, facilitating unbiased training of the model. The performance of the DNN model will be evaluated through metrics such as accuracy, precision, recall, F1-score, and the ROC-AUC score. These evaluations are crucial in determining the model's diagnostic capabilities and its potential applicability in clinical environments.
A key component of the study will be the ethical considerations, particularly focusing on the interpretability and fairness of the AI model. The project will adhere to standards that emphasize transparency and the minimization of bias in AI applications within healthcare.
Overall, the study will represent a methodical investigation into the feasibility and effectiveness of utilizing deep learning for prostate cancer screening. It aims to contribute to the broader understanding of AI's role in medical diagnostics, with a strong focus on thorough evaluation and the responsible use of technology in healthcare contexts.
To address the prevalent issue of class imbalance in medical datasets, the project will employ resampling techniques. This strategy aims to ensure a balanced representation of cancerous and non-cancerous cases, facilitating unbiased training of the model. The performance of the DNN model will be evaluated through metrics such as accuracy, precision, recall, F1-score, and the ROC-AUC score. These evaluations are crucial in determining the model's diagnostic capabilities and its potential applicability in clinical environments.
A key component of the study will be the ethical considerations, particularly focusing on the interpretability and fairness of the AI model. The project will adhere to standards that emphasize transparency and the minimization of bias in AI applications within healthcare.
Overall, the study will represent a methodical investigation into the feasibility and effectiveness of utilizing deep learning for prostate cancer screening. It aims to contribute to the broader understanding of AI's role in medical diagnostics, with a strong focus on thorough evaluation and the responsible use of technology in healthcare contexts.
Aims
- Aggregate relevant clinical and demographic data for model input;
- Construct a deep neural network specifically for prostate cancer prediction;
- Utilize resampling strategies to address imbalances between cancer and non-cancer cases;
- Evaluate the model's predictive accuracy using statistical performance metrics;
- Explore how AI can improve early prostate cancer detection;
- Assess the model's utility as a tool for supporting medical professionals;
- Provide insights into the integration of AI in cancer research and treatment strategies.
Collaborators
Julio Cesar Gali Filho