Transfer learning for grading images in photoacoustic imaging using deep learning (Conference Presentation)

Photoacoustic Imaging (PAI) is an emerging medical imaging technique that combines optical and ultrasound methods to visualize biological tissues with high contrast and spatial resolution. However, the presence of noise in PAI images often impairs their diagnostic quality. This work investigates the efficacy of transfer learning, particularly using the VGG16 model, to enhance the classification of noise types in PAI. The study addresses the challenge of limited annotated datasets in medical imaging by leveraging pre-trained convolutional neural networks (CNNs) to extract relevant features from noisy PAI images. A comprehensive approach is adopted, involving fine-tuning of the VGG16 architecture on PAI datasets to improve noise grading accuracy. Additionally, the research explores data augmentation techniques to simulate various noise conditions and thereby enhance the model’s generalization capabilities.