The PET (Positron Emission Tomography) is a functional imaging device for diagnosis of cancer, dementia, etc. And the PET-MRI (Magnetic Resonance imaging) detector is new device for medical imaging. In the PET-MRI, detector must be insensitive to magnetic field. So the PMT (PhotoMultiplier Tube) is not acceptable detector for PET-MRI. To replace PMT, researchers study about SiPM (Silicon PhotoMultiplier). SiPM has better properties in timing resolution and spatial resolution than PMT, but energy resolution is not better than PMT. To optimize the SiPM’s energy resolution, analysis of DCR (Dark Count Rate) and $\Gamma$-ray spectrum was conducted. To decide SiPM’s bias voltage, the I-V test was performed. And DCR was measured by single G-APD (Geiger mode Avalanche Photo Diode). And $\Gamma$-ray spectrum show FWHM of peak energy. The FWHM and energy resolution is signal; the DCR is noise source of energy resolution. In experiment, scintillator was LSO. And the experiments were performed by power suppliers, pre-amp board, shaping amp, and MCA (Multi Channel Analyzer). The bias voltage range is 25.5 V to 26.3 V. That range is Geiger-mode in SiPM, tested by I-V test. The DCR increases as bias voltage increases. And energy resolution decreases as bias voltage increase. The DCR was 0.3 kHz to 400 kHz. The energy resolution is about 20% ~ 22%. From Poisson calculation, noise from DCR is about 0.083 counts, noise from light statistics is about 5 counts. That means DCR has little effect to statistical noise. The energy resolution decreases as shaping time increases. But the slope is not rapid. The energy resolution is less affected by shaping time. That means the tested SiPM is in thermal noise region. In conclusion, to increase energy resolution, bias voltage increases and shaping time decreases. But bias voltage has biggest effect to energy resolution.