Purpose The gamma index analysis is a well-known approach in radiotherapy. The weakest point of widely accepted gamma index method seems to be choosing global approach. Choosing one normalization value for all points in the plan can hide potential errors, especially in its high-gradient, medium and low dose areas. In this study, the new metrics based on the local gamma analysis were introduced. Methods All analyses were performed for the TG 119 IMRT tests. The suite structures were copied to the CT scan of I’mRT Phantom (Scanditronix). Plans were prepared according to AAPM TG 119 report. All measurements were performed using Cliniac 2300C/D (Varian). The recommended point dose and plan verification by EPID were performed to confirm compliance of plans with requirements. Additional measurments with the use of ArcCHECK (SunNuclear) device and dedicated software were caried out. To perform proposed local gamma analysis the planned dose file created by TPS and the measured dose file generated by SNC software were used. Three forms of metrics were used: gamma index metric by the definition, metric proposed by Van Dyk and dose difference metric. The reference tolerance value of distance-to-agreement was set globally to 3 mm. We propose two local approaches to gamma analysis: a physics-based local approach which explicitly accounts for the physics underlying the radiation therapy process (dose value measured in local point is a random variable drawn from a Poisson distribution characterized by an expected value equal to planned dose D and standard deviation proportional to D0.5) and a clustering-based approach, which is based on clustering of the dose values read by the detectors. In the latter case two clustering methods were implemented: K-means clustering of detector readings, and Gaussian mixture clustering of detector readings. Results The planning objectives for TG 119 IMRT tests were generally met. Using proposed local approaches we were able to successfully carry out local gamma analysis of plans. We compared advantages and disadvantages of methods based on standard global and proposed local approach. Conclusions Proposed solutions have the potential to improve QA practice referring to the agreement between calculated and measured dose distributions.