AI Helps Pathologists Be More Certain Sooner

Toxicologic pathologists truly understand the need to make critical decisions with absolute confidence. Perhaps more than any other scientist, only they feel the immense pressure to reach a decision as quickly as possible.

Decisions they make on a compound during preclinical testing determine whether a drug progresses to clinical trials. The sooner this drug is pushed successfully through clinical trials, the sooner it gets to market. The sooner a medication is available, the sooner lives are saved.

AI tools applied to pathology workflows help to decrease study review times by screening high volumes of slides and providing decision support. However, barriers to AI adoption still exist, with deployment challenges that disrupt pathologist workflows. Often pathology departments in the pharmaceutical industry don’t have the time, expertise, or infrastructure for effective AI implementation.


We have helped by offering services that enable our customers to harness the power of AI at a fraction of the cost of internal deployment. We provide customers with everything needed to adopt AI immediately to solve real challenges.

Our engineers & data scientists have designed and developed AI models to address our client’s highest priorities. We provide best-in-class processing infrastructure for model development and deployment. Below are a number of examples of how we have helped our clients fast-track their projects by providing valuable diagnostic decision support to their pathologists.



Progressive pharmaceutical organisations are embracing AI’s capabilities to overcome key challenges in their work and help them make critical decisions on the safety of their drug candidates.

In a study carried out with our partner Sanofi, Patholytix AI models helped detect and quantify germinative centres in the spleen and lymph nodes in mice. Germinative centres play a key role in the body’s immune response. Different test compounds can stimulate or impede their growth which is important to quantify in drug development. Before AI, analysis of germinative centres usually required special immunohistochemistry staining, adding both time and extra cost to the process.



The goal of this project was to develop a classifier that would allow the detection of germinative centres with H&E staining only. With just a relatively low number of slides, Patholytix AI was able to build a robust classifier that accurately detected the germinative centres.

The result of this study showed a decrease in germinative centre growth in the high-dose groups, indicating the drug has an adverse effect. This project gave our partner Sanofi an alternative method to detect this vital sign of drug toxicity which is both faster and cheaper.

Often, establishing drug toxicity in an animal study can involve performing tedious tasks, such as cell or object counting. AI dramatically reduces these time-consuming tasks.


AI has been proven to increase accuracy and reduce subjectivity in drug safety assessment reviews.

In a project, we helped Janssen R&D enumerate lung macrophages in mice using Patholytix AI quantitative pathology capabilities. Macrophages are inflammatory immune cells. They exist in all tissues at basal levels. However, if present at heightened levels, this can indicate an adverse response to a drug.

In this study, a clear dose-response relationship was observed. The macrophage-specific algorithm captured both the basal level of cells within the lung tissue, but also dose-dependent changes in those levels with increasing drug administration.

In this case, AI successfully aided our partner in reaching a decision on the safety of the compound more quickly, and with greater confidence.


Another project we performed with Janssen R&D showed how AI assists in performing tasks with greater accuracy and speed.

In this project, Patholytix AI classifiers helped to enumerate necrotic neurons in mice brains. Necrosis is a form of accidental cell death caused by environmental perturbations, an example of such being drug toxicity. The neuron classifier successfully detected necrotic neurons in the area of interest, the retro spiral cortex of the brain.

Necrotic neuron counts were higher in treated animals versus control animals. Furthermore, the quantitative counts matched well with both pathologist findings and with treatment group-induced changes.


It’s natural with new technologies that there’ll be some resistance. For example, some questioned the reliability of artificial intelligence in research and clinical pathology. Our research has shown AI to be as sensitive and precise as human capabilities, if not more so.

In a study performed with Charles River Laboratories, AI was used to quantify lung neoplastic lesions in N-nitroso-N-methyl urea (NMU) treated animals. NMU is a compound often used in non-clinical studies to induce carcinogenic disease states in animals, to test the efficiency and safety of cancer treatments. Whole slide scanned lungs, thymus, and stomachs were used for supervised training.

The trained classifier was capable of identifying tumour-positive and negative animals from the positive control group with 100% concordance with the pathologists.


Some pathologists have an underlying mistrust of the application of artificial intelligence in digital pathology. They fear that it may take away the control of their diagnosis.

‘Is this going to contradict my findings?’ or even ‘‘Is this going to replace me?’.

However, those who understand the true value of digital pathology have found that it enhances their decision-making and lives as pathologists.

Working in tandem with pathologists in our partner organisations, we have shown that AI overcomes the challenges the global pathology sector faces: significant workflow inefficiencies, ageing pathologist population and declining pathologist numbers.

By offering pharmaceutical/CRO companies access to affordable outsourced AI services without investment in infrastructure, software, or personnel, our aim is to further enhance innovation in the sector. The ultimate goal is to enable our clients to accelerate drug development programs, and facilitate the journey of life-saving and life-altering drugs to the market.

To find out more about how we can assist you and your organisation fast-track your drug development projects using our AI services, please contact

Share This Post

Leave a Reply