An ΔlogR technique has a poor effect on predicting total organic carbon of source rocks with middle and deep burial depth. Aiming at this problem，on the basis of the normalization principle and one-to-one mapping，the ΔlogR technique was modified and natural gamma ray was introduced to build a model for total organic carbon prediction，by which the total organic carbon can be predicted not only in the middle and deep source rocks but also in the shallow. The introduced natural gamma data helps to reduce the impact of high resistivity of non-hydrocarbon source rocks on total organic carbon estimation and improve anti-interference ability of the prediction model. All data are normalized to solve the errors caused by different well logging instruments or standards，which also helps to reduce the complexity of data in the process.Combined with the exploration situation of southwest subsag in Dongpu Sag，including poor exploration，application of many well logging instruments，source rock buried deeply in the middle of the third member of Shahejie Formation and low acoustic value，the ΔlogR and the improved ΔlogR techniques were applied respectively to predict total organic carbon.Compared with the ΔlogR technique，the square of correlation coefficient is increased by 0.371 and the standard deviation is decreased by 0.139. It indicates that the improved ΔlogR technique could be well applied to predicting total organic carbon of source rocks with middle and deep burial depth. In addition，the improved ΔlogR technique was applied to predicting total organic carbon of source rocks in southwest subsag of Dongpu Sag，and the result indicates that the north is a potential area for petroleum exploration.