 This paper proposes a new method for detecting entanglement phase transitions in open quantum systems using a neural network decoder. The decoder is trained to predict the state of a set of reference qubits given the measurement results from the system. This allows for the detection of entanglement phase transitions even when the system size is too large to measure all the qubits directly. The authors demonstrate the effectiveness of this approach by applying it to Clifford and Harrandum circuits. They also analyze the complexity and scalability of the approach and discuss its potential applications in realistic experiments. This article was authored by Hussein Deghani, Ali Lavisoni, Muhammad Hafizi, and others.