diff --git a/docs/t4_format_3d_detailed.md b/docs/t4_format_3d_detailed.md
index b5c438d3..7d83b8ed 100644
--- a/docs/t4_format_3d_detailed.md
+++ b/docs/t4_format_3d_detailed.md
@@ -326,6 +326,7 @@ The item "description" for the category is not implemented for now.
   - "token": [str] -- Unique record identifier.
   - "name": [str] -- Category name. The latest format is "class" (e.g. car, truck), but "category.class" format (e.g. vehicle.car) is also supported.
   - "description": [str] -- Category description. Empty string `""` for now. **(Not available)**
+  - "index": [int] -- Category index, this is added to support lidarseg.
 
 For t4 format, "name" should be one of the following:
 
@@ -652,3 +653,20 @@ The instance_token is a unique identifier assigned to each object instance, allo
   }
 ]
 ```
+
+### 3D lidarseg annotation format in T4 format
+
+- T4 dataset of 3D lidarseg annotation is built upon on 3d detection of T4 dataset, the format is exactly same to [nuScenes format](https://www.nuscenes.org/nuscenes#data-format) with one additional `lidarseg.json` and lidarseg bin folder: `lidarseg/annotation/<lidarseg_token>.bin`.
+- Note that every `<lidarseg_token>.bin` consists of category index for every lidar pointcloud in a keyframe
+
+- lidarseg.json: The annotation of lidarseg for a sample data. Each record represents annotatiosn for a sample data record in T4 dataset
+
+```json
+[
+ {
+  "token":      <str> -- Unique record identifier.
+     "sample_data_token":      <str> -- Foreign key pointing to the sample data, which must be a keyframe image.
+  "filename":      <str> -- The name of the .bin files containing the lidarseg labels. These are numpy arrays of uint8 stored in binary format using numpy.
+ }
+]
+```