Summary: A groundbreaking AI language model is illuminating the complex relationship between clinical symptoms and brain tissue abnormalities. By analyzing medical summaries and tissue samples from the Netherlands Brain Bank, the model provides new insights into disease progression and the challenge of diagnosing brain diseases accurately.
This technology could significantly reduce misdiagnoses, which currently affect up to 30% of cases, by identifying molecular markers and contributing to a molecular atlas of brain disease symptoms. The ultimate goal is to improve diagnosis and open avenues for new treatments.
Key Facts:
- The AI model links clinical symptoms with brain tissue data from over 3,000 donors, offering a novel approach to understanding brain diseases.
- It identified 90 different symptoms across five domains, helping to reduce misdiagnoses by distinguishing between diseases with similar symptoms.
- The research aims to create a molecular atlas of brain diseases, which could lead to the development of targeted therapies and accurate diagnoses during a patient’s lifetime.
Source: KNAW
A new AI language model identifies clinical symptoms in medical summaries and links them to brain tissue from donors of the Netherlands Brain Bank.
This yields new insights into the development of individual disease progression and contributes to a better understanding of common misdiagnoses of brain diseases. The model may, in the future, assist in making more accurate diagnoses.
In many brain diseases, the underlying molecular mechanisms are often poorly understood, making it challenging to develop new treatment options. Investigating these molecular mechanisms is additionally challenging because the relationship between actual tissue abnormalities and the patient’s symptoms is often highly complex.
Some symptoms, for example, occur in multiple conditions, and the clinical picture can vary significantly from patient to patient, resulting in a substantial percentage of misdiagnoses (up to 30 percent). Insights gained from a newly developed AI language model may potentially change this scenario in the future.
At the Netherlands Brain Bank, brain tissue from 3,042 brain donors with a wide range of different brain diseases is stored. What makes the Netherlands Brain Bank unique is that, in addition to the tissue, they have also documented the medical history and the symptoms reported by the donors. However, this wealth of data was not quantifiable because it was transcribed in a text format, making it difficult to analyze and process.
Language Model
Inge Huitinga and her team at the Netherlands Institute for Neuroscience joined forces with Inge R. Holtman and her team at the University Medical Center Groningen to unlock this information using a new AI language model.
This classification model enables the analysis of the text in medical records and the detection of predefined symptoms. Additionally, they developed a second AI prediction model to make actual diagnoses based on the clinical picture.
Inge Holtman: ‘First, the records had to be thoroughly examined to identify symptoms that regularly occur in donors with different brain diseases.
“We eventually identified 90 different symptoms in five different domains: psychiatric symptoms (such as depression and psychosis), cognitive symptoms (such as dementia and memory problems), motor issues (such as tremors), and sensory symptoms (such as feeling things that are not there). We then manually labeled 20,000 sentences to train the classification model.’
The final model ultimately determined which symptoms occurred annually for all donors. It was observed that the prediction model was quite effective in making accurate diagnoses but fell short in rare disorders. When analyzing the diagnoses made by the prediction model, a subset of donors emerged who had been incorrectly diagnosed. It turned out that a considerable number of these donors had also been misdiagnosed by the doctor during their lifetime.
Subtypes
Holtman: ‘It seems that there is a group of people suffering from a certain condition, such as Alzheimer’s disease, but exhibiting symptoms more reminiscent of Parkinson’s disease. Or a subtype of Frontotemporal Dementia manifesting as Alzheimer’s disease. It is often challenging to diagnose these groups properly, which makes sense since these individuals show a clinical pattern that does not align with their condition. We strive to continuously improve the prediction model, hoping to make diagnoses of brain diseases more accurate.’
Inge Huitinga: ‘Understanding individual factors contributing to symptoms in brain diseases is crucial, as the reality is that many people have a combination of different conditions. Molecular markers to guide treatment are the future.
“Our ultimate goal is to develop a molecular atlas of symptoms of brain diseases. Such an atlas precisely shows which cells and molecules in the brain change with symptoms such as anxiety, forgetfulness, and depression.’
‘We expect the impact of this molecular atlas to be enormous. When we map out the molecular changes, we hope to identify the first biomarkers that can predict the correct diagnosis during a person’s lifetime. This opens doors to the development of new therapies. We are laying the foundation.’
Funding: This research is made possible by funding from the Friends Foundation from the Netherlands Institute for Neuroscience.
About this AI and neurology research news
Author: Eline Feenstra
Source: KNAW
Contact: Eline Feenstra – KNAW
Image: The image is credited to Neuroscience News
Original Research: Open access.
“Identification of clinical disease trajectories in neurodegenerative disorders with natural language processing” by Inge Huitinga et al. Nature Medicine
Abstract
Identification of clinical disease trajectories in neurodegenerative disorders with natural language processing
Neurodegenerative disorders exhibit considerable clinical heterogeneity and are frequently misdiagnosed. This heterogeneity is often neglected and difficult to study.
Therefore, innovative data-driven approaches utilizing substantial autopsy cohorts are needed to address this complexity and improve diagnosis, prognosis and fundamental research.
We present clinical disease trajectories from 3,042 Netherlands Brain Bank donors, encompassing 84 neuropsychiatric signs and symptoms identified through natural language processing. This unique resource provides valuable new insights into neurodegenerative disorder symptomatology.
To illustrate, we identified signs and symptoms that differed between frequently misdiagnosed disorders. In addition, we performed predictive modeling and identified clinical subtypes of various brain disorders, indicative of neural substructures being differently affected.
Finally, integrating clinical diagnosis information revealed a substantial proportion of inaccurately diagnosed donors that masquerade as another disorder.
The unique datasets allow researchers to study the clinical manifestation of signs and symptoms across neurodegenerative disorders, and identify associated molecular and cellular features.
