Paper-Cochlear

This repository contains code for generating and visualizing the results presented in the paper "Cochlear aqueduct advection and diffusion inferred from computed tomography imaging with a Bayesian approach".

The results for the control experiment are available in the control_results.ipynb notebook, while the results for the real data experiment are available in the real_data_results.ipynb notebook.

Below is a summary of the command-line arguments used in these two notebooks for running the Bayesian model for ear aqueduct flow model inference.

Command-Line Arguments for advection_diffusion_inference.py

This script runs the Bayesian model for ear aqueduct analysis. It has the following command-line arguments:

Key Parameters

Argument	Type	Choices	Description
-animal	str	["m1", "m2", "m3", "m4", "m6"]	The animal to model (e.g., "m1" for mouse 1).
-ear	str	["l", "r"]	The ear to model, 'l' for left or 'r' for right.
-version	str	—	User defined string used for labeling the experiment.
-sampler	str	CWMH, MH, NUTS, NUTSWithGibbs	The MCMC sampler to use.
-unknown_par_type	str	constant, smooth, step, sampleMean, custom_1, synth_diff1.npz, synth_diff2.npz, synth_diff3.npz	Type of the "true" unknown parameter (diffusion coefficient).
-unknown_par_value	str or list	—	Value(s) of unknown parameter, the diffusion coefficient, if unknown_par_type is constant, provide one value, if unknown_par_type is step or smooth, provide two values (upper and lower), if unknown_par_type is sampleMean, provide information about the samples file: tag of the experiment concatenated with the directory name where the samples are stored, separated by @
-data_type	str	real, synthetic	Type of data used for inference, real or synthetic.
-inference_type	str	constant, heterogeneous, advection_diffusion	Type of inference model, constant for assuming constant diffusion, heterogeneous for heterogeneous diffusion, and advection_diffusion for advection-diffusion model.
-Ns	int		Number of MCMC samples to draw.
-Nb	int		Number of burn-in samples.
-noise_level	str		Noise level for data, set to fromDataVar to read noise level from data that varies for each data point and set to fromDataAvg to compute average noise level from data and use it for all data points, set to avgOverTime to compute average noise level over time for each location, set to estimated to use the estimated noise level, or set to a float representing the noise level (e.g 0.1 for 10% noise). Noise level can also be a string that starts with std_ then the std value. For example std_5 means std of value 5
-add_data_pts	float list		Additional data points to add (for synthetic cases only).
-num_CA	int	(0–19)	Number of cochlear aqueduct (CA) points to use.
-num_ST	int	(0–8)	Number of scala tympani (ST) points to use.
-true_a	float		True advection speed (regarded in synthetic inference case only).
-rbc	str	zero, fromData, fromDataClip	Right boundary condition.
--adaptive	flag		Use adaptive time stepping (if passed).
-NUTS_kwargs	str		JSON-style string for NUTS sampler options (e.g. {"max_depth": 10, "step_size": 0.1}).
--data_grad	flag		If passed, gradient of concentration signal is used as data.
--u0_from_data	flag		If passed, the initial condition is obtained from the data.
--sampler_callback	flag		Enable sampler callback.
--pixel_data	flag		Treat data as pixel-level data.

note: ChatGPT was used to generate first draft of this table based on the provided code snippet of parameter definition and description.