Addressing Delicate and Variable Cancer Morphology in Spectral Histopathology Using Canine Visceral Hemangiosarcoma

Area of Science:

• Veterinary Pathology
• Biomedical Imaging
• Computational Pathology

Background:

• Spectral histopathology is effective for tumors with defined morphology.
• Application in tumors with variable or diffuse morphologies remains under-investigated.
• Canine hemangiosarcoma (HSA) presents diagnostic challenges due to its variable morphology.

Purpose of the Study:

• To evaluate Fourier transform infrared (FTIR) imaging as an accessory diagnostic tool for canine hemangiosarcoma (HSA).
• To assess FTIR's utility in classifying tumors with diffuse or variable morphologies.
• To develop and validate chemometric models for HSA diagnosis using FTIR data.

Main Methods:

• FTIR imaging was performed on canine spleen specimens without paraffin removal to preserve tissue structure.
• Partial Least Square Discrimination Analysis (PLSDA) and Random Forest (RF) models were developed for pixel-level tissue classification.
• Model performance was validated using independent test sets and a blinded set of HSA tissue samples.

Main Results:

• The PLSDA model achieved 74% correct classification of tissue types at the pixel level on an independent test set.
• FTIR spectropathology demonstrated 80% sensitivity and 81% specificity for diagnosing canine splenic HSA in blinded samples.
• The technique successfully applied to tumors with diffuse morphology, preserving delicate tissue structures.

Conclusions:

• FTIR imaging without paraffin removal is applicable to tumors with diffuse morphologies.
• This spectral histopathology approach shows significant potential as an assistive diagnostic tool for canine splenic HSA.
• The combination of FTIR imaging and chemometric algorithms offers a promising method for veterinary differential diagnosis.