Computer-aided diagnostic models to classify lymph node metastasis and lymphoma involvement in enlarged cervical...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Oncology
• Diagnostic Pathology

Background:

• Distinguishing between lymph node metastasis and lymphoma in enlarged cervical lymph nodes presents a significant clinical challenge.
• Accurate histological component identification is crucial for effective patient management.

Purpose of the Study:

• To develop and evaluate deep learning (DL)-based computer-aided diagnosis (CAD) systems for pathological diagnosis of cervical lymph nodes.
• The systems aim to differentiate between lymphomatous and metastatic involvement using positron emission tomography (PET)/computed tomography (CT) imaging.

Main Methods:

• Collected PET/CT image series from 165 patients with enlarged cervical lymph nodes.
• Constructed two DL CAD models: DL-convolutional neural networks (DL-CNN) and DL-machine learning (using DL-based features and support vector machine).
• Utilized six pre-trained CNNs for feature extraction and classification, evaluating both DL-based and handcrafted radiomics features.

Main Results:

• The DL-CNN model (ResNet50) achieved an AUC of 0.845 and 78.13% accuracy on PET/CT images.
• The DL-machine learning model (ResNet50 extractor + SVM) demonstrated superior performance with an AUC of 0.901, 86.96% accuracy, 76.09% sensitivity, and 94.20% specificity.
• Combining DL-based and handcrafted features enhanced diagnostic performance.

Conclusions:

• Developed DL-based CAD systems for classifying metastatic and lymphomatous involvement in enlarged cervical lymph nodes with high diagnostic performance.
• These systems have the potential to improve the quality of therapeutic interventions and optimize patient outcomes.