Positive identification by hand X-rays superimposition: a - PowerPoint PPT Presentation

ONE DAY FASE SYMPOSIUM SEPTEMBER 5 th , 2015 Positive identification by hand X-rays superimposition: a quantitative approach Silvio Giancola, Daniele Gibelli, Stefania Vergini, Sara Candosin, Debora Mazzarelli, Remo Sala, Cristina Cattaneo Forensic Anthropology Society of Europe – Montpellier, France – 05 Sept 2015

2 Introduction Unidentified bodies: a social emergency • Unidentified body represents 3.1% of all autopsies (Milan) • Similar percentages are reported also in USA (4.4% of unknown SUMMARY decedents every year, and • Introduction 2.6% are to become “cold cases”) • Pilot Study • • Automatic comparison 78% of cases die by traumatic • Results causes (and 22.6% by homicide) • Conclusion • Main reasons are loosening of familiar links and migration flows Positive identification by hand X-rays superimposition: a quantitative approach

3 Introduction Importance of bone features for identification Easily comparable (in some cases ID) • Prothesis and surgical devices SUMMARY • Introduction • Pathological and previous Abnormal and highly • Pilot Study traumatic lesions (bone individualizing • Automatic comparison • calluses) Results • Conclusion Modification are often • Physiological features limited and difficult to (anatomical characteristics) quantify Positive identification by hand X-rays superimposition: a quantitative approach

4 Introduction Personal identification in hands: • Looking for individual peculiarities • Comparison of similarities and differences • Recognition of people ? SUMMARY • Introduction • Pilot Study • Automatic comparison • Results • Conclusion Positive identification by hand X-rays superimposition: a quantitative approach

5 Introduction Our main goal: • Find a quantitative method for the personal identification through hand X-rays superimposition: – Find bones features that can characterize a person – Find similarities between these features in a couple of X-rays scans, – Find resemblances between a full dataset of X-rays scans, – Identify a person by its X-rays scan. SUMMARY • Introduction • Pilot Study • Automatic comparison • Results • Conclusion Positive identification by hand X-rays superimposition: a quantitative approach

6 Pilot Study • Blind tests: – 15 identifying blind tests, – 34 observers with different background, – Scores are measured in percentage of correct answers by observer. • Results: SUMMARY – Forensic Anthropologist / Odontologist : 76% of correct answers • Introduction – Anthropology Students: 67% of correct answers • Pilot Study – Forensic Pathologists: 65% of correct answers • Automatic comparison • Results • • Results are not particularly high Conclusion • Shapes analysis are done by observers Positive identification by hand X-rays superimposition: a quantitative approach

7 Automatic comparison • X-rays dataset: – 9 adults, AM, without pathologies – 2 people with 3 scans – 7 people with a single scans • Notation: X.Y , X being the person index, Y the index of the scan acquisition SUMMARY – Ex: 2.3 represent the 3 rd acquisition of the 2 nd person • Introduction • Pilot Study • Automatic comparison • Results • Conclusion Positive identification by hand X-rays superimposition: a quantitative approach

8 Automatic comparison • HALCON: Computer vision software for industrial and medical purpose – Geometrical measurements with cameras – Elaboration of 2D and 3D imagery – State of the art algorithms – Well documented examples SUMMARY – Useful tools for camera acquisition, calibration, matching, … • Introduction • Pilot Study • Automatic • Object recognition / Matching : Recognition of a model in an image comparison • Results • Conclusion – Shape-Based Matching: edge/contour detection – Deformable Matching: deformed edge/contour detection – Correlation-Based Matching: pattern/kernel recognition – Descriptor-Based Matching: keypoints/features detection Positive identification by hand X-rays superimposition: a quantitative approach

9 Automatic comparison • Segmentation of a X-rays acquisition in multiple rigid bodies (bones) SUMMARY • Introduction • Pilot Study • Automatic comparison • Results • Conclusion Positive identification by hand X-rays superimposition: a quantitative approach

10 Automatic comparison • Identification of features – Shape Based: contour identification SUMMARY • Introduction • Pilot Study • Automatic comparison • Results 2.3: 2 nd Individual - 3 rd Acquisition • Conclusion 3 rd Proximal Phalanges 2.2: 2 nd Individual - 2 nd Acquisition 2 nd Metacarpal Positive identification by hand X-rays superimposition: a quantitative approach

11 Automatic comparison • Cross comparison between scans of the dataset – Returns a matching score SUMMARY • Introduction • Pilot Study • Automatic comparison • Results 2.3 vs 2.1 • Conclusion Matching score: 99% 2.2 vs 4.1 Matching score: 73% Positive identification by hand X-rays superimposition: a quantitative approach

12 Results Ind#2 Acq#1 Metacarpals Proximal phalanges Intermediate phalanges Hand Image Individual Acquisition MC1 MC2 MC3 MC4 MC5 PP1 PP2 PP3 PP4 PP5 IP2 IP3 IP4 IP5 Score 1 Ind #1 Acq #1 72% 83% 84% 72% 65% 91% 63% 80% 87% 96% 86% 94% 95% 88% 81.87% 2 Ind #1 Acq #2 80% 70% 80% 77% 83% 92% 61% 80% 80% 97% 79% 93% 93% 93% 82.12% 3 Ind #1 Acq #3 85% 81% 96% 75% 90% 86% 63% 81% 66% 91% 69% 93% 92% 92% 82.16% 4 Ind #2 Acq #1 99% 100% 100% 100% 100% 99% 100% 99% 99% 100% 99% 98% 95% 95% 98.77% 5 Ind #2 Acq #2 91% 100% 98% 98% 99% 92% 97% 95% 94% 97% 94% 95% 99% 92% 95.74% 6 Ind #2 Acq #3 89% 98% 93% 97% 99% 94% 97% 95% 94% 99% 96% 97% 97% 91% 95.39% 7 Ind #3 Acq #1 81% 99% 89% 66% 76% 92% 79% 68% 76% 93% 67% 85% 92% 91% 81.76% 8 Ind #4 Acq #1 82% 98% 71% 69% 74% 88% 74% 70% 73% 93% 68% 89% 93% 95% 80.53% 9 Ind #5 Acq #1 84% 97% 87% 73% 75% 78% 74% 67% 70% 93% 70% 89% 97% 95% 81.40% 10 Ind #6 Acq #1 83% 91% 92% 74% 96% 76% 75% 69% 76% 93% 66% 98% 97% 95% 83.62% 11 Ind #7 Acq #1 83% 92% 92% 75% 92% 84% 75% 68% 70% 94% 70% 95% 99% 95% 83.89% 12 Ind #8 Acq #1 83% 99% 86% 74% 86% 80% 79% 73% 73% 93% 77% 89% 94% 89% 83.55% SUMMARY 13 Ind #9 Acq #1 83% 93% 88% 73% 82% 96% 84% 74% 81% 95% 77% 97% 94% 98% 86.37% • Introduction Ind#2 Acq#2 Metacarpals Proximal phalanges Intermediate phalanges Hand • Pilot Study Image Individual Acquisition MC1 MC2 MC3 MC4 MC5 PP1 PP2 PP3 PP4 PP5 IP2 IP3 IP4 IP5 Score 1 Ind #1 Acq #1 74% 91% 82% 77% 57% 76% 82% 90% 78% 88% 76% 97% 87% 81% 80.56% • Automatic comparison 2 Ind #1 Acq #2 85% 71% 92% 77% 79% 87% 77% 69% 73% 93% 75% 96% 91% 85% 81.70% • 3 Ind #1 Acq #3 88% 76% 98% 83% 88% 87% 83% 94% 80% 90% 74% 93% 86% 74% 84.98% Results 4 Ind #2 Acq #1 94% 96% 99% 98% 98% 76% 97% 97% 98% 96% 92% 98% 96% 95% 94.82% • Conclusion 5 Ind #2 Acq #2 100% 100% 100% 100% 100% 99% 100% 100% 100% 100% 100% 100% 100% 99% 99.86% 6 Ind #2 Acq #3 97% 99% 99% 97% 98% 89% 95% 97% 98% 97% 93% 99% 97% 96% 96.46% 7 Ind #3 Acq #1 73% 73% 92% 57% 52% 60% 78% 56% 64% 58% 55% 82% 77% 72% 66.82% 8 Ind #4 Acq #1 71% 71% 86% 51% 68% 51% 72% 77% 62% 52% 63% 94% 81% 79% 68.68% 9 Ind #5 Acq #1 62% 73% 91% 64% 74% 55% 58% 61% 67% 64% 67% 91% 82% 79% 69.73% 10 Ind #6 Acq #1 73% 72% 84% 66% 74% 56% 72% 87% 90% 76% 66% 96% 86% 86% 76.67% 11 Ind #7 Acq #1 72% 76% 87% 71% 76% 56% 81% 91% 92% 83% 75% 95% 89% 88% 80.15% 12 Ind #8 Acq #1 79% 76% 89% 63% 68% 56% 78% 82% 66% 80% 70% 94% 84% 72% 74.82% 13 Ind #9 Acq #1 83% 68% 94% 69% 80% 60% 83% 89% 92% 79% 72% 96% 87% 89% 80.79% Positive identification by hand X-rays superimposition: a quantitative approach