Showcasing the potential of an LLM AI integration in the digital animal patient record solution with Exolynk
Reading time: 5min
Table of contents
Abstract
With the integration of the LLM Mixtral into our digitization platform, a diagnosis of the health status of a wild animal and a treatment recommendation with medication can be created within a few seconds using existing medical parameters from the database.
Introduction
In today’s world, digitalization is more present than ever in all areas of life – and animal welfare and veterinary medicine are no exception. Wildlife sanctuaries, which are mainly funded by donations, are faced with the challenging task of finding trained veterinary nurse who are willing to work for free. In addition, large amounts of health data need to be managed in order to make the best decisions for the care of the animals taken in. This is where our solution comes in: the integration of artificial intelligence (AI) with the Mixtral 8x7b large language model in our Exolynk digitalization platform.
Application example: Digital animal patient record
Our digital animal patient record, which is available to all Exolynk users to install free of charge in our Addon Store, is already in use at several animal rescue centers in Switzerland. It enables the efficient management of animal patient records, the planning and reporting of daily care, diagnoses and treatments, an integrated eMediaktion to support the veterinary nursing staff in the selection and dosage of medication and a digital veterinary consultation for quick remote diagnosis.
The integrated AI recommendation uses the following health data to assess the state of health, indicate possible illnesses and suggest medication and dosage:
- Initial diagnosis and weight on discovery and admission to care unit
- daily weight changes
- Fecal samples and results of laboratory tests
- Daily comments from the veterinary nurses
This data is crucial for monitoring health status and the early detection and treatment of diseases.
Use of AI for health assessment
By using AI, we can generate precise assessments of health status from this data, identify potential diseases and make targeted recommendations for medication and dosage. This automated analysis enables veterinary nurse to take quick and informed action, which ultimately leads to improved animal care.
Overview of Exolynk and Mixtral 8x7b
Exolynk simplifies the digitalization of processes and the development of individual software, especially for niche applications. Thanks to drag & drop elements, predefined templates such as role models and authorization schemes, an integrated database with abstracted layers and our easy-to-learn scripting language, the development of an individual solution is around 10 times faster and more cost-effective than with conventional means and individual software.
In combination with the integration of Mixtral 8x7b, an advanced European Open Source Large Language Model (LLM) in our Script API, we offer a powerful solution for analyzing and interpreting information using generative artificial intelligence in addition to data aggregation.
Technical implementation
The key to the effective use of AI in this context lies in so-called prompt framing. We create specific queries (prompts) that enable the AI to extract relevant information from the collected data and generate an assessment of the health status and suggestions for treatment. The LLM Mixtral 8x7b used, like other models, was trained with trillions of tokens from a variety of sources, including CommonCrawl, Wikipedia, textbooks and other information available on the internet.
//Prompt Request
let question = format!("Assess the health of this hedgehog with the gender: {0} and the following initial diagnosis: {1}, the initial weight when the animal was found was: {4}g, the weight from the most recent care record entry: {6}g, the weight has changed between the last two measurements as follows: {2}g, Findings in the last fecal sample: {3}, Other comments in the care log: {5}", gender, initial_diagnosis, delta_weight, fecal_samples, entry_weight, care_remark, last_weight);
//Format & Framing Context
let llm = exo::ai::Llm::new().system_msg("You are a veterinarian who assesses the health of animals and makes a suggestion for treatment and medication in English.");
//LLM call
let answer = llm.send(question).await?;
LLM Temperature
Setting the “temperature” for the Large Language Model (LLM). The higher the value, the more creatively the LLM will answer the queries. To get an accurate result based on sound veterinary information, we set the temperature to 0.0.
//LLM Temperature
llm = llm.temperature(0.0);
This simple but powerful script demonstrates how we use Mixtral 8x7b to generate customized health recommendations from the available data.
Result of the AI recommendation
«Based on the information provided, my assessment is as follows:
The hedgehog female has shown a remarkable weight gain of 573g from the time she was found until the most recent care log entry, indicating good nutrition and care. She currently weighs 1083g, and the weight has increased by 11g between the last two measurements.
The hedgehog has lung sounds which could indicate a possible respiratory disease. Capillaria crenosome infection, caused by a parasitic worm infection, could also be a factor affecting her breathing.
My suggested treatment is to treat the hedgehog with a broad spectrum antiparasitic such as fenbendazole to eliminate the capillaria crenosome infection. At the same time, antibiotic therapy should be considered to treat any bacterial infections of the respiratory tract.
For the medication I recommend:
* Fenbendazole: 50 mg/kg body weight, orally, for 5 days
* Doxycycline: 5 mg/kg body weight, orally, for 14 days
* Prednisolone: 1 mg/kg body weight, orally, for 7 days
The hedgehog should be monitored closely to ensure that she is responding to treatment and that her condition is not deteriorating. Further veterinary consultations are recommended if her condition does not improve or worsens.
Please note that these treatment recommendations are based on the information provided and a personal examination of the hedgehog may be required to enable accurate diagnosis and treatment.»
Ethical aspects
Recommendations and assessments created using generative artificial intelligence should always be identified as such. Particularly in the medical field, such an assessment does not replace a final veterinary assessment.
Disclaimer:
Conclusion
The integration of AI into animal healthcare, especially in combination with our digital animal patient dossier on Exolynk, exemplifies the transformative power of technology. By automating data analysis, we can not only increase efficiency, but also improve the quality of animal care and counteract the shortage of skilled workers.
For local animal species in particular, for which medical literature is often only available in country-specific languages, the AI supports the veterinary nursing staff with knowledge from literary works in various languages.
We invite all interested parties to join us to discuss the possibilities, benefits and risks of this technology. Whether you work in an animal sanctuary, in the field of veterinary medicine or are simply curious about the application of AI – the Exolynk digitization platform offers the possibility to create individual digitization solutions and automate manual processes in a time and cost efficient way.
You can find more information about our digital pet patient file solution on our microsite. If you have any questions, please do not hesitate to contact us.
