Moorthy K et al, 2017: Prediction of fragmentation of kidney stones: A statistical approach from NCCT images.
Moorthy K, Krishnan M.
Lourdes Hospital, Kochi, India.
Department of Electrical and Electronics, Government Engineering College, Thrissur, S. India.
INTRODUCTION: We sought to develop a system to predict the fragmentation of stones using non-contrast computed tomography (NCCT) image analysis of patients with renal stone disease.
METHODS: The features corresponding to first order statistical (FOS) method were extracted from the region of interest in the NCCT scan image of patients undergoing extracorporeal shockwave lithotripsy (ESWL) treatment and the breakability was predicted using neural network.
RESULTS: When mean was considered as the feature, the results indicated that the model developed for prediction had sensitivity of 80.7% in true positive (TP) cases. The percent accuracy in identifying correctly the TP and true negative (TN) cases was 90%. TN cases were identified with a specificity of 98.4%.
CONCLUSIONS: Application of statistical methods and training the neural network system will enable accurate prediction of the fragmentation and outcome of ESWL treatment.
Can Urol Assoc J. 2016 Jul-Aug;10(7-8):E237-E240. doi: 10.5489/cuaj.3674. FREE ARTICLE
A method is described by means of which a statistical analysis of Hounsfield units together with a neural network system was used to predict the fragility of stones before proceeding to SWL. It is possible that such an approach may prove useful in the future. It stands to reason, however, that interpretation of this information needs to be adapted to which type of lithotripter that is used. It is surprising to find that in 120 patients as many as 73 were considered “non-fragmentable”. In the 80 cases used for training the neural network, more than half were “non-fragmentable”. It is unknown if any specific selection of patients was carried out, but this degree of disintegration is lower than generally expected.