Liver Tumor Segmentation in CT Scans Using Modified SegNet.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Hepatic cancer is a leading global cause of cancer mortality.
• Early detection through computed tomography (CT) can save lives but poses a significant workload for radiologists.
• Automated analysis of CT scans is essential for efficient and accurate diagnosis.

Purpose of the Study:

• To develop an automated method for liver segmentation and classification in CT scans.
• To address the bottleneck in processing large volumes of medical imaging data.
• To improve the speed and accuracy of hepatic cancer detection.

Main Methods:

• Adaptation and modification of the SegNet deep convolutional encoder-decoder architecture.
• Application of semantic pixel-wise classification techniques to liver CT images.
• Training and testing on a standard dataset of liver CT scans.

Main Results:

• The SegNet-based approach demonstrated high performance in liver CT segmentation.
• Achieved up to 99.9% tumor accuracy during the training phase.
• Successfully adapted a model from road scene segmentation for medical imaging tasks.

Conclusions:

• Deep learning, specifically the SegNet architecture, shows significant promise for automated liver CT analysis.
• This technique can potentially alleviate the burden on radiologists and improve early detection rates for hepatic cancer.
• Further validation on diverse datasets is warranted to confirm clinical utility.