diff --git a/include/itkCudaContextManager.h b/include/itkCudaContextManager.h
index fa3b7d3..1cd9564 100644
--- a/include/itkCudaContextManager.h
+++ b/include/itkCudaContextManager.h
@@ -43,9 +43,6 @@ class CudaCommon_EXPORT CudaContextManager : public LightObject
   static void
   DestroyInstance();
 
-  CUcontext *
-  GetCurrentContext();
-
   int
   GetCurrentDevice();
 
diff --git a/include/itkCudaImageDataManager.hxx b/include/itkCudaImageDataManager.hxx
index 57fecec..249a6ed 100644
--- a/include/itkCudaImageDataManager.hxx
+++ b/include/itkCudaImageDataManager.hxx
@@ -76,9 +76,6 @@ CudaImageDataManager<ImageType>::MakeCPUBufferUpToDate()
       std::cout << this << ": GPU->CPU data copy" << std::endl;
 #endif
 
-      CUDA_CHECK(cuCtxSetCurrent(
-        *(this->m_ContextManager->GetCurrentContext()))); // This is necessary when running multithread to bind the host
-                                                          // CPU thread to the right context
       errid = cudaMemcpy(m_CPUBuffer, m_GPUBuffer->GetPointer(), m_BufferSize, cudaMemcpyDeviceToHost);
       CudaCheckError(errid, __FILE__, __LINE__, ITK_LOCATION);
 
@@ -117,9 +114,6 @@ CudaImageDataManager<ImageType>::MakeGPUBufferUpToDate()
       std::cout << "CPU->GPU data copy" << std::endl;
 #endif
 
-      CUDA_CHECK(cuCtxSetCurrent(
-        *(this->m_ContextManager->GetCurrentContext()))); // This is necessary when running multithread to bind the host
-                                                          // CPU thread to the right context
       errid = cudaMemcpy(m_GPUBuffer->GetPointer(), m_CPUBuffer, m_BufferSize, cudaMemcpyHostToDevice);
       CudaCheckError(errid, __FILE__, __LINE__, ITK_LOCATION);
 
diff --git a/src/itkCudaContextManager.cxx b/src/itkCudaContextManager.cxx
index 4c80479..b202ab6 100644
--- a/src/itkCudaContextManager.cxx
+++ b/src/itkCudaContextManager.cxx
@@ -66,14 +66,8 @@ CudaContextManager::CudaContextManager()
   if (m_NumberOfDevices)
   {
     CUdevice device = 0;
-    m_DeviceIdx = itk::CudaGetMaxFlopsDev();
-    CUDA_CHECK(cuDeviceGet(&device, m_DeviceIdx));
-
-    CUDA_CHECK(cuCtxCreate(&m_Context, CU_CTX_SCHED_AUTO, device));
-
-    CUDA_CHECK(cuCtxSetCurrent(m_Context));
-
     m_Device = device;
+    CUDA_CHECK(cudaSetDevice(m_Device));
   }
   else
   {
@@ -83,14 +77,7 @@ CudaContextManager::CudaContextManager()
   }
 }
 
-CudaContextManager::~CudaContextManager()
-{
-  if (m_Context)
-  {
-    CUDA_CHECK(cuCtxDestroy(m_Context));
-  }
-  cudaDeviceReset();
-}
+CudaContextManager::~CudaContextManager() {}
 
 int
 CudaContextManager::GetCurrentDevice()
@@ -100,10 +87,4 @@ CudaContextManager::GetCurrentDevice()
   return device;
 }
 
-CUcontext *
-CudaContextManager::GetCurrentContext()
-{
-  return &m_Context;
-}
-
 } // namespace itk
diff --git a/src/itkCudaDataManager.cxx b/src/itkCudaDataManager.cxx
index 6db4870..850a7f0 100644
--- a/src/itkCudaDataManager.cxx
+++ b/src/itkCudaDataManager.cxx
@@ -33,9 +33,6 @@ CudaDataManager::CudaDataManager()
   // is no CUDA device available, we just do not set the context (SR). This fixes
   // the problem reported here:
   // https://www.creatis.insa-lyon.fr/pipermail/rtk-users/2015-July/000570.html
-  CUcontext * ctx = m_ContextManager->GetCurrentContext();
-  if (ctx)
-    CUDA_CHECK(cuCtxSetCurrent(*ctx));
 
   m_CPUBuffer = nullptr;
   m_GPUBuffer = GPUMemPointer::New();
@@ -91,9 +88,6 @@ CudaDataManager::Free()
   {
     try
     {
-      CUDA_CHECK(cuCtxSetCurrent(
-        *(this->m_ContextManager->GetCurrentContext()))); // This is necessary when running multithread to bind the host
-                                                          // CPU thread to the right context
       m_GPUBuffer->Free();
     }
     catch (itk::ExceptionObject & e)
@@ -171,9 +165,6 @@ CudaDataManager::UpdateCPUBuffer()
       std::cout << this << "::UpdateCPUBuffer GPU->CPU data copy " << m_GPUBuffer->GetPointer() << "->" << m_CPUBuffer
                 << " : " << m_BufferSize << std::endl;
 #endif
-      CUDA_CHECK(cuCtxSetCurrent(
-        *(this->m_ContextManager->GetCurrentContext()))); // This is necessary when running multithread to bind the host
-                                                          // CPU thread to the right context
       CUDA_CHECK(cudaMemcpy(m_CPUBuffer, m_GPUBuffer->GetPointer(), m_BufferSize, cudaMemcpyDeviceToHost));
       m_IsCPUBufferDirty = false;
     }
@@ -212,9 +203,6 @@ CudaDataManager::UpdateGPUBuffer()
       std::cout << this << "::UpdateGPUBuffer CPU->GPU data copy " << m_CPUBuffer << "->" << m_GPUBuffer->GetPointer()
                 << " : " << m_BufferSize << std::endl;
 #endif
-      CUDA_CHECK(cuCtxSetCurrent(
-        *(this->m_ContextManager->GetCurrentContext()))); // This is necessary when running multithread to bind the host
-                                                          // CPU thread to the right context
       CUDA_CHECK(cudaMemcpy(m_GPUBuffer->GetPointer(), m_CPUBuffer, m_BufferSize, cudaMemcpyHostToDevice));
     }
     m_IsGPUBufferDirty = false;