JOINT BASE PEARL HARBOR-HICKAM, Hawaii –
For many military repatriation cases, identification relies on familiar tools such as DNA, dental records, and medical radiographs. When those methods work, they provide clear and decisive answers.
But some cases fall outside that framework, especially older losses where records are incomplete or biological evidence has degraded with time. That was the situation in the case of U.S. Army Sgt. Roger Duquesne, who served with A Company, 89th Medium Tank Battalion, 25th Infantry Division during the Korean War, and whose remains had been examined repeatedly over decades without a definitive identification.
DNA testing could not advance the case because there were no family reference samples available. Dental records existed, but Duquesne had no restorations, which meant there were few distinguishing features to compare. Chest radiographs, often a powerful identification tool for Korean War-era cases, were also unavailable.
“Without family reference samples, DNA analysis often cannot move an identification forward,” said U.S. Army Col. Dori Franco, Defense POW/MIA Accounting Agency medical examiner. “Dental identification can be extremely effective, but it is relatively rare in older cases because many service members were young and had little or no dental work, and dental radiographs were not routinely collected or required at the time.”
Associate Professor Carl Stephan at the University of Queensland, who specializes in craniofacial identification methods and the video-photo superimposition technique used in this case, explained that investigators were ultimately searching for “another piece of biological evidence that could help tie things together to get it across the line for the identification.”
That missing piece came through a modern refinement of an older forensic method known as craniofacial superimposition. At its most basic level, the technique compares a photograph of a person believed to be missing with images of a skull to see whether the underlying bone structure matches the face. While the idea itself is simple, Stephan noted that the method has “always had some scientific hurdles,” particularly around subjectivity and photographic distortion. Those hurdles, he said, are what recent research has finally been able to resolve.
One of the biggest challenges is perspective. Faces look different depending on how far the camera is from the subject. “If you’re very close to someone, you get a different perspective of all of the facial features,” Stephan explained. “You must set that camera viewing the skull to be at the exact same distance from the subject as the reference face photograph, so that you can compare the features of the skull to the features of the face in a one-to-one manner.”
In Duquesne’s case, investigators were able to do exactly that because a photograph of when he was alive showed him wearing a military dress hat. The hat became a reference object. By finding a nearly identical, period-correct hat and placing it in front of a camera, investigators moved the camera backward in small, measured increments. At each step, they overlaid the live video image onto the original photograph. “You’ll find no match, no match, no match,” Stephan said, “then a sweet spot, and then no match again. That tells you how far the camera needs to be away from the reference object.”
Once that precise camera distance was identified, the hat was removed and replaced with the skull, with the camera remaining in the same position. Live video allowed the skull to be carefully rotated and positioned until it matched the pose of Duquesne’s face in the photograph. When the alignment was right, the system captured a high-resolution still image from the same position. Stephan described this hybrid approach as combining “the dynamic control of video with the precision of still photography,” something older superimposition methods could not achieve in an integrated, seamless manner.
The analysis itself focused on overall anatomical consistency rather than any single feature. The skull had to fit entirely within the outline of the face, with no bone protruding beyond the soft-tissue boundaries. Investigators examined the position of the eyes within the eye sockets, the height and shape of the nasal bridge, and the alignment of key facial landmarks. “If the skull protrudes through the soft tissue envelope,” Stephan explained, “that’s a marker for this not being a correct match.”
The teeth provided some of the most compelling evidence. In the photograph, Duquesne was smiling, showing his upper teeth. On the skull, one upper canine on the right side was particularly prominent, with the premolars behind it set slightly farther back. “That prominent canine throws a shadow across the other teeth behind it,” Stephan said. When the skull was overlaid at the correct distance and angle, the same arrangement produced the same shadow pattern seen in the photograph. “That’s one of the decisive things that indicates this is a match.”
Video-photo superimposition was not used in isolation. The results were considered alongside historical records, biological profile assessments, stable isotope analysis, and the systematic exclusion of all other possible candidates. Stephan emphasized that the method is not meant to replace DNA or dental comparison, but to support identification when those traditional tools cannot provide definitive answers. “You have to have a clear facial photograph,” he said. “The teeth need to be visible, the skull needs to be in good condition, and ideally there’s a reference object in the image. When those conditions are met, the method can be very powerful.”
For Duquesne, those conditions aligned. What had once been a stalled case moved forward because modern imaging technology allowed investigators to control perspective, reduce subjectivity, and evaluate anatomical correspondence in a scientifically defensible way. The case now stands as a clear example of how older forensic techniques can be refined rather than discarded, and how careful science, applied thoughtfully, can still bring clarity and answers to even the most difficult identifications decades after the fact.