Ovarian cancer detection using optical coherence tomography and convolutional neural networks

Area of Science:

• Biomedical imaging
• Machine learning applications in oncology
• Ovarian cancer research

Background:

• Ovarian cancer is a leading cause of cancer-related deaths in the US.
• Current non-surgical diagnostic methods lack sufficient sensitivity and specificity.
• Advancements in imaging and AI are crucial for early detection and improved patient outcomes.

Purpose of the Study:

• To develop an automated framework for ovarian cancer detection using optical coherence tomography (OCT) and machine learning.
• To evaluate the efficacy of different neural network architectures for analyzing OCT images.
• To establish a foundation for a non-surgical diagnostic tool for ovarian cancer.

Main Methods:

• Utilized optical coherence tomography (OCT) for high-resolution, depth-resolved imaging of ovaries in transgenic mice.
• Developed and compared three neural network models: VGG-supported feed-forward, 3D convolutional neural network, and convolutional Long Short-Term Memory (LSTM).
• Employed a machine learning approach for automated classification without manual feature engineering.

Main Results:

• The proposed framework successfully identified ovarian cancer in OCT recordings from transgenic mice.
• The convolutional LSTM network demonstrated superior performance, achieving a mean Area Under the Curve (AUC) of 0.81 ± 0.037.
• Models performed favorably despite inherent noise in OCT images, requiring no manual tuning.

Conclusions:

• The study presents a novel application of machine learning for analyzing depth-resolved OCT images of ovaries.
• The developed automated framework shows promise as a non-surgical diagnostic tool for ovarian cancer.
• The research has significant potential for transferability to human diagnostics, enabling earlier detection of this deadly disease.