search

thenineteen / Semiology-Visualisation-Tool

Data driven 3D brain visualisation of semiology. Semiology to anatomy translator based on over 4600 patients from 309 peer-reviewed articles.
MIT License
9 stars 6 forks source link

Bayesian average: use global-lateralisation mean between both data-subsets #224

Closed thenineteen closed 3 years ago

thenineteen commented 3 years ago

Given Bayesian posterior from TS data uses cached queries that have NEUTRAL laterality for both dominance and semiology; need a method to integrate laterality into the estimates.

When using the posterior only, this will give a neutral/symmetric result (as of c7262a1). When using SS, we can use global or micro-lateralisation. When using posterior-TS & SS average, it doesn't work as it takes the symmetric psoterior and averages it with the lateralised SS query. The localisation predictions are the mean of the two dataset estimates in order to avoid bias towards one dataset.

Need to get the posterior (TS) to give lateralised predictions. This could be done using a global-lateralisation and returning the R/L proportions to the function in getSemiologiesDataFrameFromGUI()

Then integrate the SS global lateralisation and TS-posterior global lateralisations, could use mean or inverse variance of binomial.

As the Bayesian-Average using both datasets is the mean, then we should also use the mean for global lateralisation between these datasets.

thenineteen commented 3 years ago

update: see issue #225 the option used for combining semiologies will be the same option used to combine posterior-TS and SS dataset queries