Liver Cancer Diagnosis: Enhanced Deep Maxout Model with Improved Feature Set.

Area of Science:

• Medical Imaging Analysis
• Computational Pathology
• Machine Learning in Oncology

Background:

• Accurate liver cancer classification is crucial for effective treatment planning.
• Existing classification methods often face challenges with accuracy and specificity.
• Developing robust automated systems for liver cancer detection is an ongoing research area.

Purpose of the Study:

• To propose and evaluate an advanced Deep Maxout model for liver cancer classification.
• To compare the performance of the proposed model against various established machine learning algorithms.
• To demonstrate the efficacy of the proposed scheme in improving diagnostic accuracy.

Main Methods:

• Image preprocessing using Gaussian filtering.
• LUV transformation-based adaptive thresholding for image segmentation.
• Extraction of multi-texon, Improved Local Ternary Pattern (LTP), and GLCM features.
• Classification using an improved Deep Maxout model.

Main Results:

• The improved Deep Maxout model achieved a high F-measure of 0.94.
• The proposed model outperformed Support Vector Machine (SVM), Random Forest (RF), Recurrent Neural Network (RNN), Long Short Term Memory (LSTM), K-Nearest Neighbor (KNN), Deep Maxout, Convolutional Neural Network (CNN), and other Deep Learning (DL) models in F-measure.
• The model demonstrated minimal False Positive Rate (FPR) and False Negative Rate (FNR).

Conclusions:

• The proposed liver cancer classification scheme, particularly the improved Deep Maxout model, offers superior performance.
• This advanced model provides a more accurate and reliable method for liver cancer detection.
• The findings suggest a promising direction for developing enhanced diagnostic tools in liver oncology.