We study the Roman sensitivity to exoplanets in the Habitable Zone (HZ). The Roman efficiency for detecting habitable planets is maximized for three classes of planetary microlensing events with close caustic topologies. (a) The events with the lens distances of \(D_l >\sim 7\) kpc, the host lens masses of \(M_h >\sim 0.6M_\odot\). By assuming Jupiter-mass planets in the HZs, these events have \(q\lesssim 0.001\) and \(d\gtrsim 0.17\) (\(q\) is their mass ratio and \(d\) is the projected planet-host distance on the sky plane normalized to the Einstein radius). The events with primary lenses, \(M_{\rm h} \lesssim 0.1 M_{\odot}\), while their lens systems are either (b) close to the observer with \(D_{\rm l}\lesssim 1\) kpc or (c) close to the Galactic bulge, \(D_{\rm l}\gtrsim 7\) kpc. For Jupiter-mass planets in the HZs of the primary lenses, the events in these two classes have \(q\gtrsim 0.01\), \(d\lesssim 0.04\). The events in the class (a) make larger caustics. By simulating planetary microlensing events detectable by Roman,~we conclude that the Roman efficiencies for detecting Earth- and Jupiter-mass planets in the Optimistic HZs (OHZs, which is the region between \([0.5,\sim2]\) AU around a Sun-like star) are \(0.01\%\) and \(5\%\), respectively. If we assume that one exoplanet orbits each microlens in microlensing events detectable by Roman (i.e., \(\sim 27,000\)), this telescope has the potential to detects 35 exoplanets with the projected planet-host distances in the OHZs with only one having a mass \(\lesssim 10M_{\oplus}\). According to the simulation, 27 of these exoplanets are actually in the OHZs.