PAPER ODONTOLOGY Mihran Tuceryan, 1 Ph.D.; Fang Li, 2 Ph.D.; Herbert L. Blitzer, 3 M.B.A.; Edwin T. Parks, 4 D.M.D., M.S.; and Jeffrey A. Platt, 4 D.D.S. A Framework for Estimating Probability of a Match in Forensic Bite Mark Identification* ,  ABSTRACT: In forensic dentistry, a human expert typically does the comparison and identification based on bite marks. Unlike DNA analysis, however, there is no quantitative basis with which to assign a probability for this given match. This paper proposes a framework for empirically esti- mating the probability of such a bite mark match and reports on initial experimental results. The methodology involved collection of dental popula- tion data (3D dental casts and bite mark images), image analysis for quantitatively measuring the degree of match (based on chamfer distance), and performing a logistic regression analysis using the collected population data to estimate the probability of match from the calculated degree of match. The model correctly predicted 35 of the 42 matches and 585 of the 588 mismatches. The method also has potential for use in other forensic applica- tions in which the assignment of quantitative probabilities is important. KEYWORDS: forensic science, bite mark, three-dimensional dental model, probability of match, chamfer distance, logistic regression Forensic dentistry generally addresses the problem of identifying individuals based on the properties of teeth. By looking at the loca- tion, orientation, presence ⁄ absence, and dental work, people can be matched to dental records or bite mark impressions for identification. One application of forensic dentistry, and the subject of this paper, is identifying individuals based on bite mark impressions, particularly matching bite marks left on victims to the perpetrator or matching bite marks left on criminals to the victim in instances of violent crime. It is legally relevant to match the bite mark with high confidence to the person who left it. It is also important to be able to assign a quantitative measure of the quality of the match based on objective measures. ‘‘Bite mark evidence has been almost universally accepted in the courts, but the fundamental validity and scientific basis for its use is frequently challenged (p. 1385) (1).’’ This paper focuses on the calculation of a probability of a given match being the correct one based on a goodness-of-fit measure computed geometrically and a statistical analysis of a population of three-dimensional (3D) dental casts and bite mark images generated from them. The method to compute a goodness-of-fit measure using image processing and geometric methods has been presented elsewhere (2). This paper builds upon the goodness-of-fit measure presented there and proposes a framework for developing the esti- mation of the probability measure from these results. The system is envisioned to be used in conjunction with a forensic odontologist who would be doing the actual analysis and the judgment of match. The proposed framework would then assign a quantitative probabil- ity measure for the candidate match produced by the forensic odon- tologist. In the matching experiments reported in this paper, the maxillary and mandibular anterior teeth (upper and lower, central and lateral incisors and canines) were used. Related Work There are two main applications of forensic dentistry: (i) identify- ing deceased individuals by matching antemortem and postmortem dental records and (ii) identifying criminals based on bite marks (1,3). For the first type of application, typically 2D images, such as dental radiographs and the actual teeth or their 2D images, have been used. In many situations, the identification is made manually on 2D image data by experts using software tools, such as Adobe Photoshop, to obtain the best alignment (4). Chen and Jain developed a semi-automated approach for the comparison (5,6). This method uses image registration methods applied to extracted tooth contours in potential radiographs to determine whether there is a match. In this paper, we focus on the second application, namely the identification based on bite mark impression left behind in a crime scene. Identification based on bite mark impression is made based on the shapes and arrangements of the bite mark impressions left behind and the degree of match to the teeth of the human who might have left these impressions. The bite mark impressions can be of varying qualities, and the matching of 3D teeth information 1 Department of Computer & Information Science, Indiana University Pur- due University Indianapolis (IUPUI), 723 West Michigan Street, SL 280K, Indianapolis, IN 46202. 2 Department Math and Statistics, Indiana University Purdue University Indianapolis (IUPUI), 402 North Blackford Street, LD270, Indianapolis, IN 46202. 3 Indiana Forensic Institute, 338 South Arlington Avenue, Suite 111, India- napolis, IN 46219. 4 Indiana University School of Dentistry, 1121 West Michigan Street, Indi- anapolis, IN 46202. *Supported by the US National Institute of Justice grant #2006-DN-BX- K252.   Presented at the 61st Annual Meeting of the American Academy of Forensic Sciences, February 16–21, 2009, in Denver, CO. Received 26 June 2009; and in revised form 22 Sept. 2009; accepted 10 Nov. 2009. J Forensic Sci, January 2011, Vol. 56, No. S1 doi: 10.1111/j.1556-4029.2010.01571.x Available online at: interscience.wiley.com Ó 2010 American Academy of Forensic Sciences S83