Robust alignment of prostate histology slices with quantified accuracy

Area of Science:

• Medical Imaging
• Computational Pathology
• Biomedical Engineering

Background:

• Accurate prostate tumor detection is limited by current imaging techniques.
• Precise mapping of histology data to imaged data is crucial for evaluating imaging techniques.
• 3D registration is necessary due to the misalignment between histology slices and imaging planes.

Purpose of the Study:

• To develop and validate a protocol for precise 3D registration of prostate histology slices.
• To enable accurate mapping of histology data to imaged data for improved tumor detection.
• To automate the alignment of histology slices using internal fiducial markers.

Main Methods:

• Insertion of three needles into fresh prostate specimens to create internal fiducial markers.
• Development of an algorithm for automatic detection and identification of these fiducial markers.
• Rigid alignment of histology slices based on detected fiducial markers, validated in simulated, liver, and prostate samples.

Main Results:

• Simulated images showed no residual error in the absence of deformation.
• Beef liver validation showed an average alignment accuracy of 0.12 ± 0.09 mm at fiducial markers.
• Prostate specimens achieved an average alignment accuracy of 0.18 ± 0.13 mm, with 96.1% of visible fiducial markers correctly detected.

Conclusions:

• The proposed method provides accurate and automatic 3D registration of prostate histology slices.
• The achieved accuracy is sufficient for detecting clinically significant prostate tumors.
• The protocol is efficient, requiring less than 6 minutes per specimen on average.