Are Advances in AI Technology Transforming Immunohistochemistry Analysis in Pathology?

By Dr. Chinta SidharthanReviewed by Lily Ramsey, LLMJul 3 2024

The use of artificial intelligence (AI) based tools in medicine is rapidly increasing, and novel algorithms are being created for applications in various fields of pathology to aid in and accelerate the process of analyzing pathological reports.

Immunohistochemistry is a diagnostic method that utilizes the property of antibodies to bind to their specific antigens to detect antigens in biological samples.

In a recent review published in the Journal of Personalized Medicine, researchers explored the platforms and AI-based programs that are being used to analyze immunohistochemical markers and interpret the results.

​​​​​​​Study: Revolutionizing Pathology with Artificial Intelligence: Innovations in Immunohistochemistry. Image Credit: chayanuphol/Shutterstock.com

Background

Immunohistochemistry is one of the main molecular diagnostic methods used in pathology to detect the presence of specific antigens or cellular markers in tissues, which provides pathologists and clinicians with the insights necessary to diagnose, treat, and determine the prognosis.

Over the last two decades, the development of whole-slide imaging scanners has digitized and streamlined imaging immunohistochemistry slides and the management and interpretation of the results.

The large quantities of image data acquired from these slides provide a greater complexity of information than other imaging options due to the high resolution, magnification, and color information obtained through hematoxylin and eosin (H&E) staining.

The digitized image acquisition can then be combined with various AI-based tools to analyze the images efficiently.

AI in Immunohistochemistry

In this review, the researchers provided a detailed overview of the current use of AI-based tools in analyzing immunohistochemical results. Automated AI-based analysis of images has been used to evaluate the expression of the fibroblast growth factor receptor-2 marker to understand the tumor microenvironment in breast cancer.

Furthermore, studies using machine learning to analyze breast cancer tissue imaging reported that the AI-based assessments were comparable to the manual evaluations.

Additionally, AI-based tools are also being used to analyze immunohistochemistry results quantitatively. Mindpeak, based out of Germany, has developed software modules for automated image analysis, which has been used for invasive breast carcinoma tissue samples.

A study also found that the interpretation of Ki-67 marker immunohistochemical analysis in breast cancer samples using AI tools was consistent with the gold standard and similar to the assessments made by a panel of pathologists.

AI and Immunohistochemistry-Based Cancer Diagnoses

The researchers conducted a literature review to understand the use of AI-based tools in diagnosing various cancers using immunohistochemistry.

They analyzed studies examining the use of AI-based immunohistochemical analysis in breast, prostate, and lung cancers, as well as for detecting and diagnosing malignant melanomas, lymphoid neoplasms, and cancers whose primary origins are unknown.

The researchers found that AI-based tools have not extensively been used to detect myoepithelial or basal breast cancer markers such as cytokeratin, p63, and smooth muscle actin in immunohistochemistry slides.

However, AI-based algorithms have been used to evaluate H&E-stained slides for the segmentation and classification of tumors. These algorithms can be adapted to analyze whole-slide images and detect invasive breast lesions and in-situ carcinomas.

Furthermore, several studies have investigated AI algorithms' use in immunohistochemistry to detect estrogen and progesterone receptors, as well as human epidermal growth factor receptor type 2.

The use of convolutional networks to quantify tumour-infiltrating lymphocytes in immunohistochemistry slides to detect colon, breast, gastric, and prostate cancer is also being explored.

In the case of prostate cancer, AI tools such as the machine learning-based Paige Prostate can analyze whole H&E-stained slides and classify the images as potential prostate cancer cases based on the detection of glandular atypia, or adenocarcinoma, atypical small acinar proliferation, or high-grade prostatic intraepithelial neoplasia.

Additionally, studies have also examined the use of AI-assisted analysis of multiplex fluorescence immunohistochemistry in detecting prostate tumor cells using Ki-67 quantification.

The review also discussed studies that have explored the use of convolutional neural networks to analyze H&E-stained slides for the classification of the primary subtypes of lung cancer, including small-cell carcinoma, squamous cell carcinoma, and adenocarcinoma.

AI-based tools using deep-learning algorithms have been explored to automatically measure the proliferation index, which is used for staging pigmented lesions such as malignant melanomas.

Some studies have also examined the use of AI-based tools for quantifying the immunohistochemistry markers for detecting lymphomas, the type of malignancy that involves the proliferation of lymphocytes such as T cells, B cells, and natural killer cells.

Furthermore, the researchers discussed the challenges in developing AI-based tools for detecting cancers where the primary origin of the tumor is unknown due to the scarcity of disease-specific immunohistochemistry markers for those tumors, and overlaps in the immunohistochemistry profiles with other types of cancers.

Conclusions

Overall, this comprehensive review discussed the current applications of AI algorithms and tools in streamlining and analyzing immunohistochemistry-based diagnosis of various types of cancers, especially in the assessment of H&E-stained slides.

However, the findings suggest that there are still some challenges, including socio-economic status, that prevent the widespread implementation and application of AI in the diagnostic process.