The naive approach is to self-join the table using Roads.StartPoint.STEquals(Roads.EndPoint) = 1
However, this is not a good idea and certainly won't lead to a high-performance solution. A better approach is to create a separate Nodes table that stores details of each node at which two roads intersect. This will be a single Point geometry, but you should
also assign it an integer ID.
Then, in your Roads table, add two more columns for From_NodeID and To_NodeID which reference the Nodes table, and place indexes on the columns. You can then connect roads together using a join on Roads.To_NodeID = Roads.From_NodeID, which will be more robust
and a lot faster.