We present here the first ever-reported direct computation of the cancer risk on cancer markers simultaneously detected by memristors. The novel approach in cancer diagnostics here proposed is based on the fusion of the three actions of sensing, computing and memory in a single-kind of device. So, this paper proposes the new concept of in-memory sensing as a disruptive approach in edge-computing. The concept is demonstrated by showing a well-known case in cancer diagnostics: the estimation of the risk of prostate cancer based on the simultaneous measure of Prostate Specific Antigen (PSA) and its Membrane isoform (PSMA).