Predictive Medicine for Salivary Gland Tumours Identification Through Deep Learning

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Salivary gland tumors (SGTs) are rare, with high rates of diagnostic errors complicating clinical management.
• Fine needle aspiration cytology (FNAC), the primary diagnostic tool, yields inconclusive results in approximately 25% of cases.
• There is a critical need for advanced tools to support accurate SGT diagnosis and treatment planning.

Purpose of the Study:

• To develop and evaluate a Deep Learning (DL)-based framework for automatic segmentation and classification of salivary gland tumors.
• To introduce an explainable AI approach, analyzing learning processes and attention maps to validate the framework's effectiveness.
• To enhance diagnostic accuracy and reduce errors in SGT management.

Main Methods:

• Implementation of a Deep Learning framework for automated segmentation and classification of SGTs.
• Utilizing an explainable AI methodology, including per-epoch learning analysis and attention maps.
• Evaluation of the framework on a computed tomography (CT) dataset of patients with SGTs.

Main Results:

• The proposed DL framework achieved significant performance scores in both segmentation and classification tasks for SGTs.
• The explainable AI approach provided insights into the model's learning process, supporting its reliability.
• Demonstrated potential for improved accuracy in identifying and categorizing salivary gland tumors.

Conclusions:

• The developed DL framework offers a promising solution for improving the accuracy of salivary gland tumor diagnosis.
• Explainable AI methods enhance the transparency and trustworthiness of AI-driven diagnostic tools.
• This approach has the potential to reduce diagnostic errors and guide individualized medical treatment for SGT patients.