Artificial intelligence and machine learning in ocular oncology: Retinoblastoma

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Intraocular retinoblastoma (iRB) is a serious eye cancer, particularly in children.
• Accurate diagnosis and classification are crucial for effective treatment planning.
• AI and machine learning offer potential for enhancing diagnostic accuracy in medical imaging.

Purpose of the Study:

• To evaluate the effectiveness of artificial intelligence (AI) and machine learning algorithms for diagnosing and categorizing intraocular retinoblastoma (iRB).
• To assess the performance of AI in differentiating between normal eyes and eyes with iRB, as well as classifying iRB into different stages.

Main Methods:

• A retrospective observational study was conducted.
• Artificial intelligence (AI), machine learning, and computer vision (OpenCV) were employed.
• The AI model was trained and validated on 771 fundus images from 109 eyes.

Main Results:

• The AI model achieved high sensitivity (96%) and specificity (94%) in detecting retinoblastoma (RB) across all eyes.
• Performance varied by iRB group, with high accuracy in Group A (100% sensitivity, 100% specificity) and good performance in other groups.
• The model demonstrated strong performance in classifying iRB, with high positive predictive values across most groups.

Conclusions:

• The developed AI model is highly sensitive for detecting iRB.
• The AI model exhibits high specificity for classifying iRB, supporting its utility in clinical diagnosis.
• AI shows significant potential to aid ophthalmologists in the diagnosis and management of intraocular retinoblastoma.