Visisipy: vision simulations in Python
Visisipy (pronounced /ˌvɪsəˈsɪpi/, like Mississippi but with a V) is a Python library for optical simulations of the eye. It provides an easy-to-use interface to define and build eye models, and to perform common ophthalmic analyses on these models.
Project goals
- Uniform interface to define, build and analyze various types of eye models, using abstractions that make sense in a clinical context
- Collection of ready-to-use eye models, such as the Navarro model, that can be customized at need
- Accessible interface to clinically relevant analyses on these models
Open source
The source code is available on GitHub: github.com/MREYE-LUMC/visisipy.
Contributions are welcomed! If you want to contribute, please talk to Corné Haasjes or Jan-Willem Beenakker, or email us ([email protected]).
Example: Patient-specific mapping of fundus photographs to three-dimensional ocular imaging
This notebook illustrates the ray tracing simulations used for a patient-specific mapping method of fundus photographs to three-dimensional ocular imaging on an eye model based on clinical data. Similar code written without visisipy can be found here. This project will be presented Friday at 9:45 by Jan-Willem Beenakker.
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
from helpers import InputOutputAngles
import visisipy
sns.set_style(
"ticks",
{
"axes.grid": True,
"grid.linestyle": "--",
"patch.edgecolor": "black",
},
)
geometry_parameters = {
"axial_length": 24.305, # mm
"cornea_thickness": 0.5615, # mm
"anterior_chamber_depth": 3.345, # mm
"lens_thickness": 3.17, # mm
"cornea_front_radius": 7.6967, # mm
"cornea_front_asphericity": -0.2304,
"cornea_back_radius": 6.2343, # mm
"cornea_back_asphericity": -0.1444,
"pupil_radius": 0.5, # mm
"lens_front_radius": 10.2, # mm
"lens_front_asphericity": -3.1316,
"lens_back_radius": -5.4537, # mm
"lens_back_asphericity": -4.1655,
"retina_radius": -11.3357, # mm
"retina_asphericity": -0.0631,
}
geometry = visisipy.models.create_geometry(**geometry_parameters)
model = visisipy.models.EyeModel(geometry=geometry)
field_angles = np.arange(0, 90, 5).astype(float)
raytrace_results = visisipy.analysis.raytrace(model, coordinates=zip(len(field_angles) * [0], field_angles))
fig, ax = plt.subplots()
visisipy.plots.plot_eye(ax, model, lens_edge_thickness=0.5)
sns.lineplot(data=raytrace_results, x="z", y="y", hue=[f[1] for f in raytrace_results.field], palette="plasma", ax=ax)
ax.set_aspect("equal")
ax.set_xlim(-5, 25)
ax.set_ylim(-15, 15)
ax.set_xlabel("z (mm)")
ax.set_ylabel("y (mm)")
sns.move_legend(ax, "lower right")
# Calculate cardinal point locations
cardinal_points = visisipy.analysis.cardinal_points(model)
# Get the location of the second nodal point with respect to the pupil location, which is the origin in OpticStudio
second_nodal_point = cardinal_points.nodal_points.image + (geometry.lens_thickness + geometry.vitreous_thickness)
# In the Navarro model, the second nodal point is located 7.45 mm behind the cornea apex
second_nodal_point_navarro = 7.45 - (geometry.cornea_thickness + geometry.anterior_chamber_depth)
# Calculate the location of the retina center
retina_center = geometry.lens_thickness + geometry.vitreous_thickness + geometry.retina.half_axes.axial
input_output_angles = pd.DataFrame(
[
InputOutputAngles.from_ray_trace_result(
g.set_index("index"),
np2=second_nodal_point,
np2_navarro=second_nodal_point_navarro,
retina_center=retina_center,
)
for _, g in raytrace_results.groupby("field")
]
)
fig, ax = plt.subplots()
sns.lineplot(
data=input_output_angles,
x="input_angle_field",
y="output_angle_np2",
label="$2^{\\mathrm{nd}}$ nodal point",
)
sns.lineplot(
data=input_output_angles,
x="input_angle_field",
y="output_angle_retina_center",
label="Retina center",
)
sns.lineplot(
data=input_output_angles,
x="input_angle_field",
y="output_angle_pupil",
label="Pupil",
)
ax.set_xlabel("Camera angle [°]")
ax.set_ylabel("Retina angle [°]")
ax.set_aspect("equal")
