DENTAL IMAGEPROCESSING TECHNIQUES IN FORENSIC ODENTOLOGY – A SYSTEMATIC REVIEW Rameswari Poornima Janardanan Abstract — Imageprocessing is a strong tool aiding medical and forensic research . A systematicreview is done in this paper in the area of dental image analysis applied toforensic odontology using dental x-rays. The interpretations of medical imagesrely hugely on human involvement and the human perception of the detailspresent in it.
The interpretation of thedelicate fine details in various contrast situations present in a medical imageis indeed a cumbersome task to assess. Typical radiographs obtained from aregular radiograph acquisition device may be of average quality inrepresentation. Various standardized scientific tools have been designed byresearchers, scholars and software developers to address this type ofshortcomings in a medical radiograph. These are targeted to minimize thepossible human error in predicting the right diagnosis and treatment solely onthe basis the human visual perception. Feature extraction by teeth segmentationon focused area for the information required; on an extracted tooth area in adigital dental radiograph is highlighted in this review.1. IntroductionImage processing includesseveral methods like enhancement, segmentation, filtering methods, thresholdingtechnique and morphological operations. The information from texture, shape,contours, etc is used in the classical image segmentation.
Edge detection isused to find boundaries of objects inside an image. Image enhancementtechniques are used to restore the original image. Medical imaging technology hasrevolutionized the health care over the past three decades, aiding doctors to diagnose andimprove patient outcomes. Afight against cancer is fought effectively using medical imaging in itsprevention, diagnosis and, treatment.
An important advantage of digital dentalradiography is its ability to process the image data, so that the informationcontent of the image is more accessible to the human visual system. Dentalprofessionals today are increasingly using digital dental x-rays for betterdetection, diagnosis, treatment and monitoring of oral conditions and diseases.Traditional x-ray films are replaced by the digital electronic sensors. Thesesensors can produce enhanced computer images of intra oral structures andconditions. The aim of this systematicreview is to give an overview towards current dental image processing methodsused in forensic odontology because of their potential importance in the dentaland forensic fields. Therefore,this paper is sectioned and sub sectioned as section 3.1 Reviews varioustechniques used for image segmentation and feature extraction on dentalradiographs.
Section 3.3reviews various techniques used for imagesegmentation and feature extraction on dental radiographs. This section alsohighlights the works done in forensic odontology using image segmentation.Section 3 concludes the present review. 1.1 Why it is important to do this reviewThe relevance of this review is grounded on theneed to recommend a method for dental age estimation and human identificationwith the following characteristics: simple, fast, non-invasive, non-expensive,reproducible and over all, accurate, that can be systematically used indifferent academic and forensic scenarios.
This efficiently assists in identifying deceased individuals oridentify human profiles in any doubtful situations 2. Methodology There had been many trials todevelop an automated computer vision based system to facilitate forensicodonatological applications. These systems comprises of variety of imageprocessing techniques. The basicalgorithms and methods used in dental x-ray processing are image enhancement,image segmentation, edge detection with feature extraction and neural networksbased classification.
2.1 Eligibility criteria forconsidering in this reviewThe scope of this review was notlimited to general dental image processing methods, but a brief description ofits clinical and forensic applications were reviewed. The inclusion criteria were original studies with dental image processingtechniques with forensic applications. Theeligibility criteria are as shown in the Table 1. Table 1.
The eligibility criteria for considering inthis review. Eligibility criteria Original papers Papers with different methodology Optimality of the algorithm Low failure rates Table 2. Summary of included imageprocessing methods used on dental radiographs and its purpose. Authors Image processing methods used Purpose of the study/Application Year of study Omaima Nomir, Mohamed Abdel-Mottaleb A two-stage segmentation method is used. First to separate teeth from background and the second separates upper jaw from lower jaw using integral projection. For human identification from X-ray dental radiographs 2004 Eyad Haj Said, Diaa Eldin M.
Nassar, Gamal Fahmyand Hany H. Ammar Teeth segmentation using a mathematical- morphology (MM) approach. For developing automated dental identification System 2012 A.K. Jain and H. Chen Using active contour extraction model (ACM) For matching dental x-rays for human identification 2004 R.
Cameriere, S.DeLuca, N.Egidi, M.Bacaloni, P.Maponi, L.
Ferrante, M.Cingolani. Automatic age estimation in adults by analysis of canine pulp/tooth ratio: Preliminary results To assess dental age from peri-apical x-rays 2015 Flora, G.
, Tuceryan, M. & Blitzer. Capturing the tooth contour from a bite mark image and compare it to each contour from a dental model by finding the ideal alignment and calculating goodness-of-fit. Forensic bite mark identification 2009 2.2 Exclusion criteriaStudies whichhad similar methods and often used were excluded. Non English papers were notconsidered in this review Table 2.
The list of data that was extractedfrom the reviewed full texts. Data extracted from full text items First author and Year Title Segmentation method Matching technique Purpose of the study/Application Performance rate 2.1 Study identification andselectionThe information was searched throughthe data base available through the Saudi digital library accessed through thee-library facilitated by Riyadh Colleges of Dentistry and Pharmacy. Directoryof open access journals(DOAJ), Medline/PubMed (NLM), ProQuest, Collection, (Webof science), Science Direct Journals(Elsevier), Wiley(Cross Ref),Wiley Onlinelibrary, google scholar were accessed to assimilate information this review. Reviews, articles, reports andoriginal papers published in peer journals, books, conference proceedings forgrey literature were all considered. English language publications from anysetting and recent time frame from 2010 till date, were considered eligible.The search keywords used were dentalimage processing, image segmentation on dental radiographs, humanidentification from dental x-rays, dental age estimation methods 2.2Dataextraction and managementThe collected information was organized in anexcel spread- sheet as follows: Author, year,title, enhancement technique, segmentation technique, feature extractionmethod, matching method , performance/accuracy and its applications.
2.3 Assessment of risk of bias in includedstudiesIt was necessary to avoidbias in this systematic review and thus to be free from a false positive appraisalor a false negative conclusion. The possibility of author bias was analyzed andasked for the participation of the same authors in repeated publications. The individual papers were analyzed bycomparing and then grouping them per author. Fig.
1. Flowchart of the study selection this review 3. Review of image enhancementmethods used on dental X-ray imagesIn the process of digital radiography, anelectronic senor is used to capture images of the oral cavity and itsstructures in place of traditional films. This once connected to a computer theimage can be viewed by a dentist on a screen of choice. Digital images are themost crucial medium in the field of computer vision. A digital radiograph hasthe advantage of immediate image preview and availability, and eliminates thecost of film processing steps. It provides the ability to apply special imageprocessing techniques that enhance the overall display quality of the image andextract only the regions of interests.
Withimage segmentation on digital dental radiographs, the exact information of theregion of interest can be extracted. This information is an important tool inclinical, forensic and therapeutic applications in the field of dentistry. 3.1 Reviews various techniques usedfor image segmentation and feature extraction on dental radiographsAreview on dental biometric systems and technology with further applications inforensic science was done . In2 Nomir and Adlab-Mottaleb presents a system in which, given a dental imageof a post-mortem (PM), the proposed system retrieves the best matches from anante mortem (AM) database. The system automatically segments dental X-rayimages into individual teeth and extracts the contour of each tooth.
Featuresare extracted from each tooth and areused for retrieval. During retrieval, the AM radiographs that have signaturescloser to the PM are found and presented to the user. Matching scores aregenerated based on the distance between the signature vectors of AM and PM teeth.
Figure5. Block diagram of segmentation algorithm. (Omaima Nomir ,2005) Theyintroduced iterative and adaptive thresholding. Thereafter horizontal andvertical integral projection is used for separating the jaws as well asindividual tooth. The block diagram of segmentation algorithm is as shown inFigure 5.This technique was not successful in matching images due to poorquality of images and shape of teeth could have changed with time as PM imageswere taken after a long time AM images were captured.
In 3, Eyad Haj Said, Diaa Eldin M Nassar, Gamal Fabry & HanyAmmar presented method of teeth a mathematicalmorphology approach to the problem of teeth segmentation. They also proposed agrayscale contrast stretching transformation to improve the performance ofteeth segmentation. We The various techniques for dental segmentation of X- ray imagesaddress the problem of identifying each individual tooth and how the contoursof each tooth are extracted is presented. Their technique was not able toproperly segment an X- ray by a single segmentation technique and it variedfrom image to image. (Said,E.
H,2006) in his paper designed anapproach based on mathematicalmorphological segmentation. Greyscale contrast stretching transformation isperformed for an enhanced teeth segmentation performance. It presented atechnique with a low failure rate on comparison to other approaches. Figure 6 Mainstages of the algorithm(Said,E.H,2006)Figure . 7Grayscale line profiles of the input image, the upper horizontal line profileillustrates the bones between the teeth, the lower horizontal line profileshows the gap between the teeth, while the vertical line profile illustratesthe gap valley. (Said,E.H,2006) In5, Hong Chen & A.
K. Jain introduced dental biometrics using activecontour extraction model (ACM). As per this paper traditional snake cannot ableto discriminate edges of multiple adjacent objects. So there can be presence ofoverlapping images.
To remove this problem the authors utilized directiongradients. This proposed system has main two stages: feature extraction,matching. In this to extract contours of dental work the intensity histogram ofthe tooth image is automated with the mixture of Gaussian model. In thematching stage three steps given: Tooth level matching, tooth contours arematched using a shape registration method, and the dental work is matched onoverlapping areas. Distance between postmortem and ante mortem radiographsprovide candidates identities to estimate subject identification. The toothcontour is the feature extracted as they remain invariant over time incomparison to other feature of the teeth.
Radiograph segmentation and contourextraction are done in the feature extraction stage. Based on edge detectioncontour extraction is approached. Figure 9 .Theprocessing flow diagram(Chen and Jain 2005) andthe results of teeth alignment and dental work alignment with the parametersused in teeth alignment. (a) Query DW. (b) Genuine DW.
(c) Imposter DW. (d) Thecontours of the DW in (b) and the DW in (a) being affine transformed with theteeth alignment parameters between (a) and (b). (e) The contours of the DW in(c) and the DW in (a) being affine transformed with the teeth alignmentparameters between (a) and (c). (Chen and Jain 2005) In their paper R. Cameriere, S.
DeLuca in 2015 proposed for the first time automatingteeth segmentation for the purpose of dental age estimation based on a previouslyproposed formula. Here the segmentation is done in two steps for the tooth andthe pulp.As the intensity of the pixels is greater in the tooth than in thebackground, a suitable threshold is selected. From the knowledge of set T, apiecewise linear approximation of its boundary is computed. This is animportant item of information in the segmentation of the pulp. In fact, theboundary of the tooth encloses the region where the pixels of the pulp area canbe found.
Shape analysis is applied to all the transversal sections of thetooth. The shape analysis of the transversal sections is based on thecharacteristic M-shape of the corresponding grey level function. In everytransverse section, the arithmetic mean of the grey levels of pixels around themiddle of the pulp area are compared to the arithmetic mean of the grey levelsof the pixels around the boundary of the pulp area.
When the absolute value oftheir difference is lower than a given threshold it is the pulp end. Finally, apiecewise linear approximation of the points of the pulp boundary is obtainedby a least square polyline of these points Flora, G., Tuceryan in 2009 proposed a method for bite markidentification by extracting tooth contour for matching. The general steps forbite mark identification are as follows: 1. create a digitization of the set of3-dimensional dental casts. 2. capture a 2-dimensional contour of the teethfrom each digitization. 3.
capture the tooth contour from each bite mark image.4. compare each bite mark image contour to each contour from the dental modelby finding the ideal alignment and calculating goodness-of-fit. 5. thecomparison which causes the maximum goodness of-fit is identified as the match Fig: 11.Typical captured tooth contours using the deformable curve. Fig 12.
Fig 12. Matching resultsfor the comparison using bite mark contours extracted with deformable curves. Table 3. Acomparison between the teeth segmentation algorithms Algorithm Principles Type of view Is it automated Jain and Cheng Integral projection Bitewing & OPG Semi-Automated Nomair &Abel Motleb Iterative &adaptive Thresholding, integral projection Bitewing Automated E .H.
Said, D. Nassar Mathematical morphology of teeth Bite wing and periapical Automated R. Cameriere, &S.DeLuca, Thresholding technique and shape analysis Periapical Automated Flora, G.&Tuceryan, M. A 2-dimensional contour of the teeth from 3-dimensional dental casts Semi automated In teeth segmentation algorithms optimality andpercentage of failure shows light on performance of segmentation algorithmsMeasuring suboptimality measures the performance of algorithms in between thetwo extremes. In practical cases, it is difficult to achieve optimalperformance with 100% images, and when comparing segmentation algorithms, oneshould favor those whose failure rates are the lowest and their optimality andlow-order measures of suboptimality predominate the testing results 16.
Thefailure rate is especially important when assessing teeth segmentationalgorithms, since those films where no teeth can be properly segmented cannotbe used in the identification process.4. Results and conclusionFor feature extraction andsegmentation most research scholars makes use of thresholding and morphologicaloperation. From the review of abovepapers, the main challenge in developing an automated dental recognizetionsystem is to deal with poor quality of images, imaging angle, teeth overlap,teeth shape change matter due to aging, occluded teeth, etc. Incorporating artificialintelligence(AI) tools such as neuralnetworking, fuzzy c-means, etchas not been much explored, for the better understanding and diagnosis purpose.Researchers up till now have beenfound concentrating on image enhancement or segmentation forextracting features for forensic sciences for human identification. No deepresearch has been published for automating dental atlas for human ageprediction for odonatological purposes. Bite mark analysis using imageprocessing is still not well explored.
Automated or semi-automated diagnosis of aforesaid objectives would be quite useful for further identification ofhuman and especially for asylum seeker issues. This systematic review summarizes and compares the results of some of the most used methods for dental imagesegmentation methods used in forensic application mainly for humanidentification. In the light of the evidence one could identify there is a needto identify a teeth segmentation algorithm with a better performance rate forfurther forensic and clinical applications