The role of artificial intelligence in healthcare: Personalized treatments, diagnostic support, and prediction of animal movements – Image: Xpert.Digital
Transformation through AI in the body & cosmos: How algorithms heal heart defects & count whales
AI as a key technology in healthcare and species conservation: Artificial intelligence as a game changer
Artificial intelligence (AI) is no longer just a buzzword from science fiction films, but a reality that permeates our lives in countless ways. Particularly in healthcare and species conservation, AI is unlocking enormous potential, revolutionizing traditional methods and opening up entirely new avenues. We are at the dawn of an era in which AI not only serves as a supporting tool, but also acts as a driving force for innovation and progress. This report highlights how AI is already making a crucial difference in three key areas—personalized treatment of atrial fibrillation, AI-assisted diagnostics in digital pathology, and predicting animal movements to protect marine ecosystems—and promises even greater transformation in the future.
Suitable for:
Personalized treatment of atrial fibrillation through AI: A paradigm shift in cardiology
Atrial fibrillation, the most common sustained heart rhythm disorder, affects millions of people worldwide and places a significant burden on healthcare systems. Treating this complex condition is often challenging, as its course can vary considerably from patient to patient. This is where AI comes in, enabling a fundamental shift towards personalized treatment approaches.
AI-optimized ablation procedures: Precision and effectiveness at a new level
One particularly promising area is catheter ablation, a minimally invasive procedure for treating atrial fibrillation. This method involves selectively destroying diseased heart tissue that causes the arrhythmia. Traditionally, ablation was often performed using a rather standardized, anatomically oriented approach. However, the TAILORED-AF trial, a milestone in interventional cardiology, has demonstrated how AI can significantly improve the precision and effectiveness of this procedure.
In this randomized, controlled trial, a subset of patients underwent AI-based technology called Volta AF-Xplorer™. This system analyzed over 5,000 data points per second in real time during the procedure and identified spatiotemporally dispersed electrograms—a complex pattern of electrical signals indicative of pathological areas of the heart muscle. Compared to the control group, which underwent ablation using conventional methods, the AI-assisted cohort showed impressive results. After 12 months, 88% of patients in the AI group were free of arrhythmias, compared to only 70% in the control group. Furthermore, acute recurrences occurred significantly less frequently in the AI group (15% vs. 66%). These results demonstrate that AI is capable of processing enormous amounts of data intraoperatively during ablation, enabling more precise and individualized treatment.
The term "ablation" comes from Latin and means "to take away" or "to remove." In medicine, it describes the targeted removal or destruction of tissue. Besides catheter ablation for cardiac arrhythmias, there are numerous other applications, such as tumor ablation, in which tumor tissue is destroyed using heat, cold, or other methods, or endometrial ablation, which is used to treat certain gynecological conditions. Catheter ablation has established itself in recent years as one of the most important treatment options for atrial fibrillation and is now becoming even more effective and safer thanks to AI-assisted procedures.
Predictive models for treatment success: risk profiles and personalized prognoses
Another promising approach in the field of AI-assisted atrial fibrillation therapy is the development of predictive models. The ACCELERATE project, led by the Leipzig Heart Center, is working on machine learning models that can create individual risk profiles based on 12-lead ECG data. These models go far beyond simply predicting the recurrence of atrial fibrillation after ablation. They are also able to detect left atrial remodeling—a fibrotic remodeling process of the left atrium that not only promotes the development of atrial fibrillation but is also associated with a significantly increased risk of stroke. Studies show that left atrial remodeling can increase the risk of stroke by 3.2 times.
To maximize the predictive accuracy of these models, registry data from over 100,000 ablations (as of 2021) are integrated. The results are impressive: The models achieve a predictive accuracy of 89% for so-called low-voltage areas in the heart, i.e., areas with reduced electrical activity that often correlate with fibrotic tissue. Compared to conventional risk scores used in clinical practice, the AI-based models outperform them by 23%. This means that AI is able to identify patients who have a particularly high risk of recurrent atrial fibrillation or stroke, thus enabling personalized treatment planning. In the future, such predictive models could help physicians choose the optimal treatment strategy for each individual patient and thus maximize treatment success.
Pulsed-field ablation (PFA): The next generation of ablation technology
In addition to optimizing existing ablation techniques, AI is also driving the development of entirely new methods. One example is pulsed-field ablation (PFA), an innovative technology that uses electrical pulses to selectively destroy heart muscle cells. Unlike conventional ablation methods based on heat or cold, PFA uses ultrashort, high-frequency electrical fields. This results in highly targeted necrosis of the heart muscle cells while sparing surrounding tissue, such as the esophagus or the phrenic nerve.
AI plays a crucial role in PFA by adapting the pulse rate to tissue thickness in real time. This ensures optimal ablation effect with maximum safety. Initial studies at the German Heart Center Berlin (DHZC) show promising results. For example, the procedure time was reduced by up to 40% using PFA compared to conventional ablation methods. At the same time, the procedure demonstrated a high level of safety, particularly regarding the protection of the esophagus and phrenic nerve, which can sometimes be damaged during conventional ablation procedures. PFA could therefore make atrial fibrillation ablation not only more efficient but also safer, and the treatment more comfortable for patients.
AI in digital pathology and diagnostic support: Precision and speed in the service of diagnosis
Pathology, the study of diseases, plays a central role in medical diagnostics. Traditionally, pathological diagnostics is based on the microscopic examination of tissue samples. This process is time-consuming, subjective, and can be affected by human fatigue and variability. Digital pathology, the digitization of tissue sections and the use of computer-aided analysis methods, promises a revolution in this area. AI is a key factor in fully utilizing digital pathology and raising diagnostics to a new level.
Automated tumor detection: Identifying cancer cells with deep learning
A key application of AI in digital pathology is automated tumor detection. The Fraunhofer Institute for Microelectronic Circuits has developed deep learning algorithms that can identify malignant cell clusters in digitized tissue sections with impressive precision. These algorithms have a sensitivity of 97%, meaning they correctly detect tumor cells in 97% of cases.
By employing transfer learning, a machine learning method that transfers knowledge from one task to another, the system was trained on a massive database of 250,000 histopathological images. This enables the system not only to recognize tumor cells but also to differentiate between 32 subtypes of ductal carcinoma, the most common form of breast cancer. This detailed subtyping is crucial for treatment planning. Furthermore, the AI can reduce diagnostic time in pathology by up to 65%, leading to faster diagnoses and thus earlier initiation of therapy for patients. Automated tumor detection using AI can therefore significantly improve the efficiency and accuracy of pathological diagnostics while simultaneously reducing the workload for pathologists.
Neural networks in routine pathology: Detecting overlooked micrometastases
Another example of the successful use of AI in pathology is the work of the company Aisencia, which employs convolutional neural networks (CNNs). These specialized neural networks are particularly adept at recognizing patterns in images and are used in digital pathology to predict, for example, microvascular invasion in colon cancer. Microvascular invasion, the penetration of tumor cells into the smallest blood vessels, is an important prognostic factor in colorectal cancer and provides information about the risk of metastasis.
In a validation study of 1,200 samples, Aisencia's AI achieved a 94% agreement with the assessments of experienced pathologists. This demonstrates that the AI is capable of detecting microvascular invasions with a similar level of accuracy to human experts. Remarkably, however, the AI in this study also detected an additional 12% of micrometastases that were missed during the initial assessment. This underscores the potential of AI to recognize subtle patterns and details that might escape the human eye. The use of CNNs in routine pathology can therefore improve the quality of diagnostics and help ensure that no important information is overlooked.
SATURN: AI-based diagnosis of rare diseases – Putting an end to diagnostic odysseys
Rare diseases pose a particular challenge to the healthcare system. Often, years pass before patients with a rare disease receive the correct diagnosis. These so-called "diagnostic odysseys" are very stressful for those affected and their families. AI can make a significant contribution here by accelerating and improving the diagnostic process.
The smart physician portal SATURN is an example of an AI-based system that combines Natural Language Processing (NLP) with knowledge graphs to generate differential diagnoses from symptom lists. NLP enables the AI to understand and process natural language, while knowledge graphs represent medical information and relationships in a structured format. In the pilot phase of the project, SATURN was tested for the diagnosis of rare metabolic disorders. The system correctly identified 78% of cases of Gaucher disease and 84% of mucopolysaccharidoses. The misclassification rate was only 6.3%.
A particular advantage of SATURN is its connection to SE-ATLAS, a directory of specialized treatment centers for rare diseases. This allows the system not only to support diagnosis but also to directly suggest suitable experts and centers. This can significantly shorten the time to the correct diagnosis and treatment. Studies show that SATURN can reduce the average diagnosis time from 7.2 years to 1.8 years. AI-based diagnostic support systems like SATURN have the potential to fundamentally improve the care of patients with rare diseases and spare them unnecessary suffering.
Predicting whale movements using AI-supported satellite analysis: Species conservation in the 21st century
AI is playing an increasingly important role not only in healthcare but also in species conservation. Monitoring and protecting endangered animal species are crucial for preserving biodiversity. Traditional methods of animal observation are often time-consuming, expensive, and difficult to cover large areas. AI-supported satellite analysis and acoustic monitoring open up entirely new possibilities for efficiently and comprehensively recording animal movements, thus making species conservation more effective.
SPACE WHALE: Deep Learning for Marine Megafauna – Counting Whales from Space
The SPACEWHALE system, developed by BioConsult SH, is a striking example of how AI and satellite technology can be combined to monitor marine megafauna. SPACEWHALE analyzes satellite images with an extremely high resolution of 30 cm (provided by Maxar Technologies) using an ensemble of CNNs and random forest models. These AI models are trained to detect and classify whales in satellite images.
In Auckland Bay, a key habitat for southern right whales (Eubalaena australis), SPACEWHALE was successfully deployed. The AI detected 94% of the whales present in the area. Manual validation by experienced marine biologists confirmed the system's high accuracy of 98.7%. SPACEWHALE reduces the cost of whale surveys by up to 70% compared to traditional aerial counts. Furthermore, the method enables, for the first time, large-scale population surveys in the open ocean, areas that are difficult to access using conventional methods. SPACEWHALE demonstrates how AI-powered satellite analysis can revolutionize species conservation by providing more precise, cost-effective, and widespread monitoring capabilities.
Acoustic monitoring and habitat modeling: Hearing whales and predicting migration routes
In addition to visual monitoring using satellite imagery, acoustic monitoring also plays a crucial role in species conservation. The WHALESAFE project off the coast of California combines hydrophone data (underwater microphones) with AI-based LSTM (Long Short-Term Memory) networks to predict the presence of blue whales in real time. LSTM networks are a special type of neural network that excels at recognizing temporal relationships in data.
In addition to acoustic data, the WHALESAFE models also consider environmental factors such as sea temperature, chlorophyll A concentration (an indicator of algal blooms and thus food availability), and shipping traffic data. By combining these diverse data sources, the models achieve an impressive 89% accuracy rate in predicting blue whale migration routes. A key objective of WHALESAFE is to reduce ship collisions, one of the main threats to whales. Automatic warnings to ships entering critical areas have already reduced the collision rate in the Santa Barbara Channel by 42%. WHALESAFE demonstrates how AI-powered acoustic monitoring and habitat modeling can contribute to better protecting whales and other marine life and minimizing human-wildlife conflict.
Real-time detection of communication signals: Understanding the language of sperm whales
A particularly fascinating and forward-looking project in the field of AI-supported species conservation is the Cetacean Translation Initiative (CETI). CETI aims to decipher the communication of sperm whales. Sperm whales are known for their complex click sounds, known as "codas," which they use to communicate with each other. The CETI project analyzes over 100,000 hours of sperm whale clicks using Transformer models. Transformer models are a state-of-the-art neural network architecture that has proven particularly powerful in natural language processing in recent years.
Through contrastive learning, a machine learning method in which AI learns to distinguish between similar and dissimilar data points, CETI's AI recognizes context-specific codas. These codas are used, for example, in coordinating dives or raising young. Initial results suggest that sperm whale communication has a syntax with recurring five-element sequences. These findings could provide insights into intentional communication, meaning that sperm whales are capable of communicating consciously and purposefully with one another. CETI is an ambitious project that could not only revolutionize our understanding of whale communication but also open new avenues for species conservation by enabling us to better address the needs and behaviors of these fascinating animals.
Key technology for a better future
The examples in this report vividly demonstrate that the integration of AI into healthcare and species conservation is already having a transformative impact. In cardiology, AI enables more precise and personalized ablation procedures; in pathology, it accelerates and improves tumor diagnostics; and in species conservation, it is revolutionizing the monitoring of marine species and allowing for a deeper understanding of complex animal behavior. But this is just the beginning.
Future fields like quantum machine learning, which could harness the immense computing power of quantum computers, promise further breakthroughs in arrhythmia prediction and other medical areas. In species conservation, swarm intelligence-based systems that replicate the collective behavior of insect or bird swarms could be used for whale tracking and the protection of entire ecosystems. However, to fully exploit the potential of AI-driven innovations, close interdisciplinary collaboration between medicine, computer science, ecology, and many other disciplines is essential. Only through the exchange of knowledge and expertise can we ensure that AI technologies are used responsibly and for the benefit of both people and the environment. The future is intelligent – let's shape it together.
Suitable for:
Your global marketing and business development partner
☑️ Our business language is English or German
☑️ NEW: Correspondence in your national language!
Konrad Wolfenstein
I would be happy to serve you and my team as a personal advisor.
You can contact me by filling out the contact form or simply call me on +49 89 89 674 804 (Munich) . My email address is: wolfenstein ∂ xpert.digital
I'm looking forward to our joint project.