diff --git a/Project.toml b/Project.toml
index 1efc826..ddebd55 100644
--- a/Project.toml
+++ b/Project.toml
@@ -9,6 +9,6 @@ Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 
 [compat]
-LazyGrids = "0.4, 0.5"
+LazyGrids = "0.5"
 SpecialFunctions = "1.8, 2"
-julia = "1.9"
+julia = "1.10"
diff --git a/README.md b/README.md
index 96c79fe..de899ec 100644
--- a/README.md
+++ b/README.md
@@ -135,7 +135,7 @@ ellipsoids/spheres, cuboids/cubes, gaussians, cylinders, cones.
 
 ### Compatibility
 
-Tested with Julia ≥ 1.9.
+Tested with Julia ≥ 1.10.
 
 <!-- URLs -->
 [action-img]: https://github.com/JuliaImageRecon/ImagePhantoms.jl/workflows/CI/badge.svg
diff --git a/src/mri-sense.jl b/src/mri-sense.jl
index e49347f..0b4612b 100644
--- a/src/mri-sense.jl
+++ b/src/mri-sense.jl
@@ -29,7 +29,7 @@ leading to better boundary behavior than the DFT frequencies `k/N`.
 * `kmax::Int = 5` default frequency index -kmax:kmax in all dimensions
 * `kfun::Function = (k,N) -> k / (2N)` # DCT-II frequency
 * `deltas::NTuple{D,<:Number} = ones(D)` pixel sizes
-(For additional options `kmaxs`, `kt`, `ki`, see code.)
+(For additional options `kmaxs`, `kt`, `ki`, `T`, see code.)
 
 # Output
 * `(; B, ν)` where `B` is basis matrix of size `count(mask) × nk`
@@ -47,7 +47,7 @@ function mri_smap_basis(
     kfun::Function = (k,N) -> k / (2N), # DCT-II frequency
     nfun::Function = (N) -> -(N÷2):(N÷2)-1,
     basis::Function = (k,N) -> cispi.(2 * nfun(N) * kfun(k,N)), # 1D basis
-    T::DataType = ComplexF32,
+    T::Type{<:Complex{<:AbstractFloat}} = ComplexF32,
 ) where D
     B = zeros(T, count(mask), length(ki))
     for (i, k) in enumerate(ki)
diff --git a/src/object2.jl b/src/object2.jl
index 990086e..a13d16f 100644
--- a/src/object2.jl
+++ b/src/object2.jl
@@ -86,7 +86,7 @@ function phantom(
     y::AbstractVector,
     oa::Array{<:Object2d},
     oversample::Int ;
-    T::DataType = typeof(1f0 * oneunit(promote_type(eltype.(oa)...))),
+    T::Type{<:Number} = typeof(1f0 * oneunit(promote_type(eltype.(oa)...))),
 )
 
     oversample < 1 && throw(ArgumentError("oversample $oversample"))
diff --git a/src/object3.jl b/src/object3.jl
index cc205b9..09447c4 100644
--- a/src/object3.jl
+++ b/src/object3.jl
@@ -90,7 +90,7 @@ function phantom(
     z::AbstractVector,
     oa::Array{<:Object3d{S,V}},
     oversample::Int;
-    T::DataType = typeof(oneunit(V) * 1f0), # at least Float32
+    T::Type{<:Number} = typeof(oneunit(V) * 1f0), # at least Float32
 ) where {S, V <: Number}
 
     oversample < 1 && throw(ArgumentError("oversample $oversample"))
diff --git a/src/shepplogan.jl b/src/shepplogan.jl
index 755f47a..d7c6061 100644
--- a/src/shepplogan.jl
+++ b/src/shepplogan.jl
@@ -93,7 +93,7 @@ Convenience method for generating `M×N` samples of Shepp-Logan phantoms.
 # Options
 * `oversample::Int = 3` (usually)
 * `yflip::Bool = true` (reverse `y` samples for convenience.)
-* `T::DataType` default `Float32` (except `Int` for `BrainWeb` version)
+* `T::Type{<:Number}` default `Float32` (except `Int` for `BrainWeb` version)
 * `kwargs...` remaining options passed to `ellipse_parameters` for parameters.
 
 # Out
@@ -107,7 +107,7 @@ function shepp_logan(
     case::EllipsePhantomVersion ;
     oversample::Int = 3,
     yflip::Bool = true,
-    T::DataType = Float32,
+    T::Type{<:Number} = Float32,
     kwargs...
 )
     ob = shepp_logan(case ; kwargs...)