Exploring Machine Learning and Deep Learning for Early Detection of Prostate Cancer
Abstract
Prostate cancer remains one of the most prevalent cancers among men, and improving early detection is critical for reducing morbidity and mortality. This research-in-progress explores multiple publicly available prostate cancer datasets to evaluate how artificial intelligence can support earlier risk identification and diagnostic decision-making. Rather than focusing on a single dataset or modeling approach, this project intentionally investigates several potential data directions, including structured clinical datasets (e.g., PLCO), large-scale inpatient databases (e.g., HCUP), surgical registries, and imaging collections such as ProstateX MRI. The first methodological direction centers on tabular clinical modeling, analyzing variables such as age, BMI, smoking status, alcohol consumption, diet type, physical activity level, family history, stress level, sleep duration, routine health screenings, PSA level, DRE results, biopsy outcomes, and urinary symptoms. Classical machine learning models, including logistic regression, random forests, support vector machines, and gradient boosting, will be used to estimate cancer risk and evaluate feature importance, supporting interpretable early detection tools. A second direction explores the use of language models and structured clinical data integration to analyze patient-level features and risk patterns, with the goal of understanding how modern AI systems can assist in risk categorization and clinical insight generation. Longer-term expansion includes image-based deep learning using prostate MRI and histopathology data, where convolutional neural networks may detect subtle patterns not captured in tabular analysis. Overall, this project evaluates multiple AI pathways, from interpretable machine learning to deep multimodal architectures, to identify scalable, responsible approaches for prostate cancer early detection.
Start Time
15-4-2026 9:00 AM
End Time
15-4-2026 10:00 AM
Room Number
272
Presentation Type
Oral Presentation
Presentation Subtype
Research-in-Progress
Presentation Category
Science, Technology, and Engineering
Faculty Mentor
Husari Ghaith
Exploring Machine Learning and Deep Learning for Early Detection of Prostate Cancer
272
Prostate cancer remains one of the most prevalent cancers among men, and improving early detection is critical for reducing morbidity and mortality. This research-in-progress explores multiple publicly available prostate cancer datasets to evaluate how artificial intelligence can support earlier risk identification and diagnostic decision-making. Rather than focusing on a single dataset or modeling approach, this project intentionally investigates several potential data directions, including structured clinical datasets (e.g., PLCO), large-scale inpatient databases (e.g., HCUP), surgical registries, and imaging collections such as ProstateX MRI. The first methodological direction centers on tabular clinical modeling, analyzing variables such as age, BMI, smoking status, alcohol consumption, diet type, physical activity level, family history, stress level, sleep duration, routine health screenings, PSA level, DRE results, biopsy outcomes, and urinary symptoms. Classical machine learning models, including logistic regression, random forests, support vector machines, and gradient boosting, will be used to estimate cancer risk and evaluate feature importance, supporting interpretable early detection tools. A second direction explores the use of language models and structured clinical data integration to analyze patient-level features and risk patterns, with the goal of understanding how modern AI systems can assist in risk categorization and clinical insight generation. Longer-term expansion includes image-based deep learning using prostate MRI and histopathology data, where convolutional neural networks may detect subtle patterns not captured in tabular analysis. Overall, this project evaluates multiple AI pathways, from interpretable machine learning to deep multimodal architectures, to identify scalable, responsible approaches for prostate cancer early detection.
