Segmentation of chondroblastoma from medical images using modified region growing algorithm P. Y. Muhammed Anshad 1 • S. S. Kumar 2 • Shajeem Shahudheen 3 Received: 8 December 2017 / Revised: 8 January 2018 / Accepted: 31 January 2018 Ó Springer Science+Business Media, LLC, part of Springer Nature 2018 Abstract Chondroblastoma is a bone tumor typically found within the cartilage tissue space of bone and unfolds into its sur- roundings. This tumor reduces the strength of bone and even leads to death if not treated early. Chondroblastoma is diagnosed from X-ray images and the tumor can be removed by surgical methods. For the successful removal, the exact volume of tumor should be known, which can be identified by segmenting the tumor region from the image. Chon- droblastoma can be segmented from X-ray image by manual and computer aided methods. Manual segmentation may leads to inter and intra observer errors. This work proposes an efficient segmentation tool called modified region growing method for segmenting chondroblastoma. This work focuses on automatic and accurate segmentation of chondroblastoma which gives better segmentation results than existing methods. The segmentation results are evaluated using dice coefficients, jaccard distance, and coefficient of similarities, spatial overlaps, absolute volume measurement error and figure of merit. Keywords Computer  Aided diagnosis  Modified region growing algorithm  Volume measurement 1 Introduction Bone diseases are thought of seriously due to bones importance to the life of persons. Many distinct varieties of tumours or lesions can develop within the bone and Chondroblastoma is the one type generally affects the epiphyses or epiphyses of long bones. It arises from an outgrowth of immature cartilage cells (chondroblasts) from secondary ossification centres, originating from the epi- physeal plate or some remnant of it. Chondroblastoma is typically found in early age below 25 years and also the correct diagnosis in early stage will cure the patient utterly. Computer-aided diagnosis that focuses on computerized analysis of medical images is employed by radiologists as a ‘‘Second Opinion’’ in detecting lesions, assessing extent of sickness, creating diagnostic selections and convenient volumetry [1]. Typically, this has been manually done by trained clinicians. Manual segmentation is time intense, user dependent, error prone and needs skilled information to yield correct and sturdy results. Even once well-trained consultants perform this operation were employed using manual segmentation is susceptible to errors related to inter observer and intra observer variability. Alternatives to manual segmentation are planned employed using a type of computer-assisted strategies. Segmentation of Chondrob- lastoma from X-ray images is a particularly difficult task due to the variations in size and form of bone across patients, the presence of tumours on boundaries, and dif- ferent abnormalities [2]. X-ray is commonly used mainly due to cost effectiveness. Different approaches were used for segmentation of bone tumor and lesion from X-ray images. The intensity threshold based segmentation methods need some thresh- old values to be fed manually, the deformable model techniques want sizable amount of manual inputs, statisti- cal model-based methods [3–6], the look for best model and requires an excessive amount of time, level set needs high computation, and Snake algorithm [7–11] needs goodish processing time. Fuzzy c means [12] based lesion segmentation is not effective due to noisy or outlying & P. Y. Muhammed Anshad mohamednool78967@gmail.com 1 Department of Electronics & Communication, Noorul Islam University, Kanyakumari, India 2 Electronics and Instrumentation, Noorul Islam University, Kanyakumari, India 3 Vivid Scans, Cochin, India 123 Cluster Computing https://doi.org/10.1007/s10586-018-1954-0