Improving the diagnosis accuracy is essential for the clinical application of osteoporosis evaluation using ultrasonic backscatter signal.In vitro ultrasonic backscatter signals were measured on bone specimens and backscatter parameters were calculated.Using the measured backscatter parameters,the involved cancellous bone specimens were evaluated and classified using support vector machine and adaptive boosting algorithms.Results showed that the accuracy of classification was 80.00％-82.86％ and the specificity of osteoporosis diagnosis was significant (specificity ＞ 92.3％).The supervised machine learning method using ultrasonic backscatter in bone evaluation is effective in the diagnosis of osteoporosis.The performance of the proposed machine-learning method is superior to the traditional bone evaluation using quantitative backscatter parameters.This study may contribute to the application of ultrasonic backscatter in the diagnosis of osteoporosis in vivo.