Fisher encoding of convolutional neural network features for endoscopic image classification.

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Accurate diagnosis of celiac disease (CD) and colonic polyps (CP) is crucial for patient outcomes.
• Current diagnostic methods can be time-consuming and may require extensive training data.

Purpose of the Study:

• To develop and evaluate an automated approach for diagnosing CD and CP using convolutional neural network (CNN) activations and Fisher encoding.
• To assess the performance of the proposed method against existing CNN- and non-CNN-based techniques.

Main Methods:

• Applied three CNN architectures (AlexNet, VGG-f, VGG-16) to endoscopic image databases for CD and CP.
• Utilized Fisher encoding on convolutional layer activations, classified with support vector machines.
• Experimented with concatenating Fisher representations from multiple layers.

Main Results:

• The proposed CNN-Fisher approach demonstrated superior performance compared to other methods.
• The method achieved high accuracy without requiring training on the specific target endoscopic dataset.
• Significant time savings were observed due to the elimination of target dataset training.

Conclusions:

• The CNN-Fisher approach offers a highly effective and efficient solution for automated diagnosis of CD and CP.
• This method reduces computational time and resources by eliminating the need for dataset-specific training.
• The approach holds promise for improving the speed and accuracy of gastrointestinal disease diagnosis.