Artificial intelligence: its use in medical diagnosis

Area of Science:

• Nuclear Medicine
• Artificial Intelligence
• Medical Imaging

Background:

• Artificial neural networks (ANNs) offer advanced capabilities for image processing and pattern recognition.
• The application of ANNs in nuclear medicine is being explored for analysis, interpretation, and diagnosis.
• Emerging hardware developments are enabling larger and more complex neural network systems.

Purpose of the Study:

• To evaluate the effectiveness of ANNs in classifying normal versus abnormal FDG-PET scans.
• To assess the utility of ANNs in interpreting ventilation/perfusion (V/Q) scans for pulmonary embolism diagnosis.
• To explore the potential of ANNs in medical imaging applications.

Main Methods:

• An artificial neural network (ANN) was employed to classify FDG-PET scans.
• ANNs were utilized to interpret data from standard V/Q scans.
• The performance of ANNs was compared against traditional methods like discriminant analysis.

Main Results:

• The ANN demonstrated superior performance compared to discriminant analysis in classifying FDG-PET scans.
• Favorable results were achieved using ANNs for interpreting V/Q scan data to determine the likelihood of pulmonary embolism.
• The development of hardware neural systems, such as an electronic retina, suggests future advancements.

Conclusions:

• ANNs are a promising technology for various applications in nuclear medicine, including image analysis and diagnosis.
• The successful application of ANNs in FDG-PET and V/Q scan interpretation highlights their diagnostic potential.
• Advancements in ANN hardware are expected to drive further innovation and widespread adoption in medical imaging.