diff --git a/modules/nvidia_plugin/src/cuda/cuda_type_traits.hpp b/modules/nvidia_plugin/src/cuda/cuda_type_traits.hpp
index a8a231dd9..9e01b77e3 100644
--- a/modules/nvidia_plugin/src/cuda/cuda_type_traits.hpp
+++ b/modules/nvidia_plugin/src/cuda/cuda_type_traits.hpp
@@ -5,6 +5,7 @@
 #pragma once
 
 #include "openvino/core/type/element_type.hpp"
+#include "openvino/core/type/element_type_traits.hpp"
 
 #ifdef __CUDACC__
 #include <cuda/float16.hpp>
diff --git a/modules/nvidia_plugin/src/ops/convert_color_i420.hpp b/modules/nvidia_plugin/src/ops/convert_color_i420.hpp
index f9c4e9602..3f5386fc6 100644
--- a/modules/nvidia_plugin/src/ops/convert_color_i420.hpp
+++ b/modules/nvidia_plugin/src/ops/convert_color_i420.hpp
@@ -31,9 +31,9 @@ class I420ConvertColorBase : public OperationBase {
         constexpr const size_t N_DIM = 0;
         constexpr const size_t H_DIM = 1;
         constexpr const size_t W_DIM = 2;
-        NGRAPH_CHECK(node.get_input_size() == 1 || node.get_input_size() == 3,
-                     "I420 conversion shall have one or 3 inputs, but it is ",
-                     node.get_input_size());
+        OPENVINO_ASSERT(node.get_input_size() == 1 || node.get_input_size() == 3,
+                        "I420 conversion shall have one or 3 inputs, but it is ",
+                        node.get_input_size());
         auto single_plane = node.get_input_size() == 1;
 
         const auto& in_tensor_shape = node.get_input_shape(0);
diff --git a/modules/nvidia_plugin/src/ops/convert_color_nv12.hpp b/modules/nvidia_plugin/src/ops/convert_color_nv12.hpp
index 20168f6d7..5a160ad71 100644
--- a/modules/nvidia_plugin/src/ops/convert_color_nv12.hpp
+++ b/modules/nvidia_plugin/src/ops/convert_color_nv12.hpp
@@ -31,9 +31,9 @@ class NV12ConvertColorBase : public OperationBase {
         constexpr const size_t N_DIM = 0;
         constexpr const size_t H_DIM = 1;
         constexpr const size_t W_DIM = 2;
-        NGRAPH_CHECK(node.get_input_size() == 1 || node.get_input_size() == 2,
-                     "NV12 conversion shall have one or 2 inputs, but it is ",
-                     node.get_input_size());
+        OPENVINO_ASSERT(node.get_input_size() == 1 || node.get_input_size() == 2,
+                        "NV12 conversion shall have one or 2 inputs, but it is ",
+                        node.get_input_size());
         const bool single_plane = node.get_input_size() == 1;
 
         const auto& in_tensor_shape = node.get_input_shape(0);
diff --git a/modules/nvidia_plugin/src/transformer/nodes/fully_connected.cpp b/modules/nvidia_plugin/src/transformer/nodes/fully_connected.cpp
index 70c7aaa04..915eb70da 100644
--- a/modules/nvidia_plugin/src/transformer/nodes/fully_connected.cpp
+++ b/modules/nvidia_plugin/src/transformer/nodes/fully_connected.cpp
@@ -78,7 +78,7 @@ ov::PartialShape validate_matmul_output_shape(const ov::PartialShape& arg0_shape
     auto arg0_rank = arg0_shape.rank().get_length();
     auto arg1_rank = arg1_shape.rank().get_length();
 
-    NGRAPH_CHECK((arg0_rank != 0 && arg1_rank != 0), "Scalars are not supported as MatMul inputs.");
+    OPENVINO_ASSERT((arg0_rank != 0 && arg1_rank != 0), "Scalars are not supported as MatMul inputs.");
 
     // Temporary Dimension vectors to calculate output shape
     std::vector<ov::Dimension> arg0_shape_tmp(arg0_shape);
@@ -117,17 +117,17 @@ ov::PartialShape validate_matmul_output_shape(const ov::PartialShape& arg0_shape
     auto merged_dimension = ov::Dimension::dynamic();
     auto arg0_col_dim = arg0_shape_tmp[arg0_rank - 1];
     auto arg1_row_dim = arg1_shape_tmp[arg1_rank - 2];
-    NGRAPH_CHECK(ov::Dimension::merge(merged_dimension, arg0_col_dim, arg1_row_dim) || arg0_col_dim.is_dynamic() ||
-                     arg1_row_dim.is_dynamic(),
-                 "Incompatible MatMul matrix dimension. ",
-                 "First input dimension=",
-                 arg0_col_dim,
-                 " at COL_INDEX_DIM=",
-                 (arg0_rank - 1),
-                 " doesn't match the second input dimension=",
-                 arg1_row_dim,
-                 " at ROW_INDEX_DIM=",
-                 (arg1_rank - 2));
+    OPENVINO_ASSERT(ov::Dimension::merge(merged_dimension, arg0_col_dim, arg1_row_dim) || arg0_col_dim.is_dynamic() ||
+                        arg1_row_dim.is_dynamic(),
+                    "Incompatible MatMul matrix dimension. ",
+                    "First input dimension=",
+                    arg0_col_dim,
+                    " at COL_INDEX_DIM=",
+                    (arg0_rank - 1),
+                    " doesn't match the second input dimension=",
+                    arg1_row_dim,
+                    " at ROW_INDEX_DIM=",
+                    (arg1_rank - 2));
 
     // 3. If ranks of input arguments are different after steps 1 and 2,
     // the smaller tensor is unsqueezed from the left side of the shape
@@ -152,15 +152,15 @@ ov::PartialShape validate_matmul_output_shape(const ov::PartialShape& arg0_shape
             // to ensure MatMul backward compatibility.
             // Instead fully dynamic dimension is set as default for such a case.
             auto merged_dimension = ov::Dimension::dynamic();
-            NGRAPH_CHECK(ov::Dimension::merge(merged_dimension, arg0_shape_tmp[i], arg1_shape_tmp[i]) ||
-                             arg0_shape_tmp[i].is_dynamic() || arg1_shape_tmp[i].is_dynamic(),
-                         "Incompatible MatMul batch dimension. ",
-                         "Can't merge first input dimension=",
-                         arg0_shape_tmp[i],
-                         " with second input dimension=",
-                         arg1_shape_tmp[i],
-                         " at index=",
-                         i);
+            OPENVINO_ASSERT(ov::Dimension::merge(merged_dimension, arg0_shape_tmp[i], arg1_shape_tmp[i]) ||
+                                arg0_shape_tmp[i].is_dynamic() || arg1_shape_tmp[i].is_dynamic(),
+                            "Incompatible MatMul batch dimension. ",
+                            "Can't merge first input dimension=",
+                            arg0_shape_tmp[i],
+                            " with second input dimension=",
+                            arg1_shape_tmp[i],
+                            " at index=",
+                            i);
 
             output_shape[i] = merged_dimension;
         } else {