Cell orientation entropy (COrE): predicting biochemical recurrence from prostate cancer tissue microarrays

Area of Science:

• Digital pathology
• Computational biology
• Cancer imaging analysis

Background:

• Biochemical recurrence (BCR) is a critical endpoint for prostate cancer (CaP) patients post-surgery.
• Accurate prediction of BCR is essential for effective patient management.
• Current methods may not fully capture subtle morphological changes indicative of recurrence.

Purpose of the Study:

• Introduce Cell Orientation Entropy (COrE), a novel feature descriptor for cancer cell analysis.
• Quantitatively model the disorder of cell/nuclear orientation in local neighborhoods.
• Evaluate the correlation between directional disorder measurements and BCR in CaP patients.

Main Methods:

• Developed COrE to rigorously quantify cell/nuclear orientation and local cell networks.
• Utilized second-order statistical features to measure disorder in local cell orientation.
• Applied 39 COrE features to CaP tissue microarray (TMA) images for BCR prediction.
• Employed a random forest classifier with 3-fold cross-validation.

Main Results:

• Achieved an accuracy of 82.7 +/- 3.1% in predicting 10-year BCR using a combination of COrE and other nuclear features.
• Demonstrated the ability of COrE features to capture characteristics of cell orientation relevant to disease progression.
• The model was evaluated on a dataset comprising 19 BCR and 20 non-recurrence patients.

Conclusions:

• COrE is a novel and effective feature descriptor for digital pathology image analysis.
• Measurements of directional disorder using COrE show significant correlation with BCR in prostate cancer.
• COrE features hold potential for characterizing disease states in various histological cancer images beyond prostate cancer.