Evaluation of Diagnostic Accuracy of Machine Learning Algorithm for Computer Aided Diagnosis (CAD) of Space Occupying Lesions in Computed Tomography (CT) Images of Brain

Abstract

Objectives-Objective of this project was to formulate a low cost and userfriendly SOL detection computational model through machine learning and test its diagnostic efficacy through retrospective evaluation of radiologist confirmed normal and abnormal CT images. Materials & Methods-The neural network has been created in the Wolfram language, using our own dataset. A total of 249 images were obtained from a CT scan center’s human brain data. A database of 89 normal and 76 SOL images from 128 slice CT scan machines was used for formulating the model. The test trial was done using 50 different normal and 34 different abnormal images. Results-The sensitivity of ability to identify abnormal images was 97% and the specificity was 92%. Only one out of 34 images with tumor was reported wrongly and 4 normal images out of 50 normal test images were wrongly identified by the CAD system. The software was able to achieve a diagnostic accuracy higher than the confidence interval of 95% set for the diagnostic test. Conclusion - The CAD system is novel and innovative as it is a simple, lowcost and user-friendly computational model which can be used by doctors on their office PCs.

Authors and Affiliations

Rushank Goyal, Swati Goyal, Lovely Kaushal

Keywords

Related Articles

OSTEOSYNTHESIS OF SUBTROCHANTERIC FRACTURES USING PROXIMAL FEMORAL NAIL: A PROSPECTIVE STUDY

Background Subtrochanteric fractures remain one of the most...

Postoperative Complications In Operated Cases Of Pyogenic Spondylodiscitis

Study Design: A prospective study from January 2013 to December 2015 was conducted at a single tertiary-care institution. Objective - The aim of the study was to identify the postoperative adverse events and characterize...

Functional outcome of minimally invasive plate osteosynthesis of proximal tibial fractures using locking compression plates

Introduction: Aim of surgical treatment of proximal tibial fractures is to restore and preserve normal knee function. Locking compression plate device offers potential bio-mechanical advantages. Treatment of these injuri...

Evaluation Of The Efficacy Of Cytology And Colposcopy In Diagnosing Cervical Lesions : A Clinical Study Sippy Agarwal, Saurabh Ag

Carcinoma cervix is commonest genital malignancy in India in women. It is a preventable carcinoma. Various screening methods are available. This study is done in 200 women to show the comparision between cytology and...

Ejaculation Preserving Trans Urethral Resection Of Prostate (epTURP): An Initial Study

A short term prospective study was conducted on 10 patients (age range 55-65 years and sexually active) undergoing the ejaculation preserving Trans Urethral Resection of Prostate (epTURP). They were evaluated for preserv...

Download PDF file
  • EP ID EP540139
  • DOI -
  • Views 108
  • Downloads 0

How To Cite

Rushank Goyal, Swati Goyal, Lovely Kaushal (2018). Evaluation of Diagnostic Accuracy of Machine Learning Algorithm for Computer Aided Diagnosis (CAD) of Space Occupying Lesions in Computed Tomography (CT) Images of Brain. International Journal of Orthopaedics Traumatology & Surgical Sciences, 4(2), 218-222. https://europub.co.uk/articles/-A-540139