Pushing the boundaries of digital pathology with AI
Histopathology is a diagnostic discipline that involves the microscopic examination of tissue samples in health and disease. It plays a critical role in medical diagnostics, helping to understand the structural and cellular changes in tissues that occur due to various diseases. Instead of examining tissue samples on glass slides under a microscope, pathologists nowadays very often use high-resolution digital images that are captured, viewed and analysed on a computer. These images can be shared remotely with other pathologists, significantly enhancing diagnosis and patient care through access to expert opinions and diagnoses.
Incorporating AI in digital pathology
Undertaken with the support of the Marie Skłodowska-Curie Actions programme, the CLARIFY project developed a robust automated digital diagnostic environment based on advanced AI techniques. The aim was to enhance the interpretation of histological images in a digital setting. AI has emerged as a transformative force in medical diagnostics, particularly in the field of pathology. By leveraging complex algorithms and machine learning, AI can analyse vast amounts of medical data with unprecedented speed and accuracy. “AI's ability to detect subtle patterns and anomalies in histological images promises to enhance diagnostic accuracy, reduce variability among pathologists, and ultimately improve patient outcomes,” emphasises project coordinator Valery Naranjo.
Web platform for pathologists
The CLARIFY project initially focused on three challenging cancer types: triple-negative breast cancer, high-risk non-muscle invasive bladder cancer, and Spitzoid melanoma. However, it has the potential to extend to other types of cancer by utilising publicly available databases. One of the standout achievements of the CLARIFY project is the design of a prototype web platform for pathologists. The CLARIFY platform aims to use a decentralised data flow management architecture for histological images, seamlessly integrating essential analysis tools within a federated cloud environment. Moreover, the platform features a model for content-based histopathological image retrieval, which allows users to upload an image and receive the most similar case in response, thereby aiding pathologists in diagnosing complex cases. Although the platform’s clinical readiness is yet to be verified, the content-based histopathological image retrieval methodology has been validated.
Digital pathology innovations
To standardise digital pathology assessment and reduce variability among pathologists, CLARIFY has developed customised technical solutions. With AI models, it has made it possible to objectively assess clinicopathologic parameters and enhance tumour classification and histopathological interpretation. Specifically, the project used cutting-edge deep neural network models and tailored architectures for feature extraction and classification. CLARIFY introduced diagnostic and prognostic classification pipelines, which utilise sophisticated algorithms to analyse histological images and accurately classify various types of cancers. These pipelines assist pathologists in not only identifying the presence of cancer, but also in predicting the likely course and outcome of the disease. This has the potential to assist in treatment planning and patient management.
Training network
CLARIFY set up a training network for early-stage researchers. The consortium successfully coordinated a diverse group of students from various backgrounds, including medicine and engineering, to advance digital pathology practices. “Moving forward, we are looking for more funding opportunities for further development and improvement of the histopathological diagnostic process,” concludes Naranjo.
Keywords
CLARIFY, AI, cancer, digital pathology, histopathology