Prostate Cancer Diagnosis via Visual Representation of Tabular Data and Deep Transfer Learning.

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Prostate cancer (PC) is a significant global health concern for men.
• Current diagnostic methods like biopsies and digital rectal examination (DRE) have limitations including invasiveness, cost, and accuracy.
• There is a need for improved, non-invasive diagnostic tools for PC.

Purpose of the Study:

• To develop and evaluate a novel machine learning approach for prostate cancer diagnosis.
• To explore the application of various machine learning techniques, including deep learning and transfer learning, on tabular clinical data.
• To assess the effectiveness of a new Tab2Visual framework for transforming tabular data into graphical representations for analysis.

Main Methods:

• Utilized traditional, tree-based, and advanced tabular deep learning methods for PC data analysis.
• Adapted convolutional neural networks (CNNs) and transfer learning for tabular data using the proposed Tab2Visual framework.
• Developed ensemble models to further enhance prediction accuracy.

Main Results:

• The proposed machine learning approach achieved a high F1-score of 0.907.
• The model demonstrated strong performance with an Area Under the Curve (AUC) of 0.911.
• Experimental evaluation confirmed the effectiveness and superiority of the developed method.

Conclusions:

• The novel machine learning approach shows significant promise for accurate prostate cancer detection.
• This method offers a potential non-invasive and cost-effective alternative to traditional diagnostic procedures.
• The Tab2Visual framework successfully enables the application of image-based deep learning techniques to tabular data for medical diagnosis.