Insights and resources for AI in pathology

Case study: AI image analysis reveals novel findings in Lewy Body Disease

Written by Aiforia | Oct 10, 2025 9:23:14 AM

Dr. Liisa Myllykangas, Associate Professor of Neuropathology and Consultant Neuropathologist from the University of Helsinki and Helsinki University Hospital, Dr. Ville Kivistö, and the neuropathology research team developed an artificial intelligence-based algorithm to quantify tyrosine hydroxylase (TH)-immunoreactive sympathetic distal axons at the myocardial level.

This article summarizes the study and is written based on the full publication and an interview with Dr. Myllykangas. Read the complete publication from Acta Neuropathologica: Myocardial sympathetic distal axon loss in subjects with Lewy pathology in three autopsy cohorts

Lewy Body Disease (LBD) = a neurodegenerative disease characterized by the accumulation of misfolded alpha-synuclein (α-syn), which clinically manifests as Parkinson’s disease or dementia with Lewy bodies. The pathological hallmark of LBD is Lewy pathology, which consists of Lewy bodies and neurites found in neurons.

 

Hypothesis – where it all started?

The team hypothesized that an AI model quantifying tyrosine hydroxylase (TH)-immunoreactive distal axons at the myocardial level would provide a sensitive method to study the possible myocardial sympathetic denervation/dysfunction in subjects with different forms of synucleinopathy. 

“AI has quickly become a valuable tool in pathological research. In our case, we hypothesized AI would provide a novel way to accurately quantify distal axons of sympathetic nerves in myocardial samples (this has proven to be difficult using conventional microscopy). Compared to the alternative of conventional microscopy, using Aiforia significantly sped up the analysis of the samples and provided statistically powerful quantitative data.” – Liisa Myllykangas, M.D., Ph.D, Associate Professor of Neuropathology and Consultant Neuropathologist, University of Helsinki and Helsinki University Hospital

By applying the AI model to three independent autopsy cohorts, the team wanted to show that CNS Lewy pathology is strongly associated with loss of distal sympathetic axons at the myocardial level. This is the first histopathological evidence supporting the body-first (caudo-rostral) subtype of LBD being prone to earlier and more severe myocardial denervation/dysfunction, occurring independently of other known causes of sympathetic denervation/dysfunction.

 

From sample collection to analysis


Samples were collected from three different cohorts

The following independent autopsy cohorts were included in the study:

  • Vantaa 85+ consists of samples from individuals 85 years old or older living in the city of Vantaa (southern Finland) on April 1, 1991. Heart samples were collected from the septum in a standardized manner, regardless of macroscopic pathology, for 138 neuropathologically examined subjects. 
  • The Helsinki Biobank includes samples acquired as part of routine autopsies performed at the Helsinki University Hospital. In total, samples from 87 cases were used in this study: 54 cases neuropathologically diagnosed with LBD, whose DLB consortium classification was defined, 13 cases neuropathologically diagnosed with MSA, and 20 cases that had not been neuropathologically diagnosed with LBD.
  • The Tampere Sudden Death Study (TSDS) is a forensic autopsy cohort, collected by Prof. Pekka Karhunen and Dr. Eloise Mikkonen, Tampere University, consisting of 700 cases that died outside a hospital in the Tampere region (southern Finland) in 2010–2015. For this study, the team acquired left ventricle samples from 37 cases previously shown to have Lewy pathology, 86 age and sex-matched controls without Lewy pathology, and 4 cases with MSA-type pathology, for a total of 127 subjects. 


In all cohorts, cases with CNS Lewy pathology formed the “LP-positive” group, and those without CNS Lewy pathology formed the “LP-negative” group. MSA cases were considered separately as the “MSA” group.

To learn more about further subtyping and staining, review the complete publication

Developing the AI model

To analyze the sympathetic innervation in the heart samples, the team developed an image recognition algorithm using Aiforia® Create. It was trained to quantitatively measure the area of TH-reactive staining in digitalized cardiac tissue sections, measure the tissue area, and separate larger nerve fascicles from other tissue. Samples from all three cohorts were used in the training, and the performance of the AI model was validated by comparing its results against annotations made by three board-certified pathologists.

“Using Aiforia® Create was intuitive, so even someone without extensive experience with machine learning and programming could use it effectively with the help of the online resources by Aiforia. On top of that, our contacts with the Aiforia team assisted with the planning of this project and provided additional help with the more advanced parts of the Aiforia® Platform when needed,” Dr. Myllykangas describes. 

 

Novel findings on a myocardial level

In all three autopsy cohorts, the presence of Lewy pathology was strongly associated with loss of TH reactivity at the myocardial level. However, Dr. Myllykangas explains that the most significant finding of the study was that in the Vantaa 85+ autopsy-cohort, the caudo-rostral (or “body-first”) subtype of Lewy body disease was associated with significantly decreased cardiac sympathetic innervation, whereas the amygdala-based (or “brain-first”) subtype was not. “This is the first study to report this finding at the myocardial level. Our finding provides a histopathological verification for the findings of a previous imaging study based on living patients,” she describes. 

In conclusion, this study's results show that Lewy pathology in the central nervous system, and particularly its caudo-rostral subtype, is strongly associated with loss of sympathetic distal axons at the myocardial level. As described above, the team also provided evidence that the caudo-rostral subtype is one of the strongest predictors of myocardial sympathetic denervation/dysfunction in the oldest-old population.

“This project was a great learning experience for our group, and we are planning to use AI in our future projects,” Dr. Myllykangas concludes. 

 

Explore more Aiforia® Create use cases →