Image processing for pathological visualization in multitemporal convoluted TIRI
Daniel T. J. Arthur,Masood Mehmood Khan,Luke C. Barclay +2 more
- 08 Jun 2012
- pp 725-728
TL;DR: In this article, a longitudinal dataset of clinical thermal infrared images was processed to facilitate visualization of osseous stress pathology in the lower limbs of Australian Army basic trainees, and the use of thermal chroma-keying in segmentation and multitemporal image calibration was demonstrated.
read more
Abstract: The convoluted nature of thermal infrared radiation and poor understanding of the physical mechanisms of human emittance, make objective image acquisition and processing protocols prerequisite for meaningful diagnostic specificity. A longitudinal dataset of clinical thermal infrared images was objectively processed to facilitate visualization of osseous stress pathology in the lower limbs‥ This paper details processing of 500+ thermal infrared images acquired during a recent three month clinical study into osseous stress pathology in the lower limbs of Australian Army basic trainees. The use of thermal chroma-keying in segmentation and multitemporal image calibration is demonstrated. The ‘OpenSURF’ implementation of the scale and rotation-invariant interest point detector and descriptor are shown to be performant in registration of multitemporal clinical thermal infrared image data. Thermal ‘signs’ observed in longitudinal images appear to be revealing detectable changes in osseous stress pathophysiology.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Figures
![Fig. 1. Intertemporal image calibration offset taken from the centroids of the histogram representation of the segmented, emissively uniform, backgrounds of images taken from the same recruit during weeks 1, 4, 7, and 9 of their basic training, as detailed in [5].](/figures/fig-1-intertemporal-image-calibration-offset-taken-from-the-2ml8jzaj.png)
Fig. 1. Intertemporal image calibration offset taken from the centroids of the histogram representation of the segmented, emissively uniform, backgrounds of images taken from the same recruit during weeks 1, 4, 7, and 9 of their basic training, as detailed in [5]. 
Fig. 3. Detection of interest points in a pre-processed TIRI (a) via the fast Hessian detector approximation (b) in OpenSURF. Detected interest points are plotted onto (a) in (c). ![Fig. 2. Preprocessed multitemporal TIRI sequence from a symptomatic participant [5].](/figures/fig-2-preprocessed-multitemporal-tiri-sequence-from-a-26kixx96.png)
Fig. 2. Preprocessed multitemporal TIRI sequence from a symptomatic participant [5]. ![Fig. 8. Cadaveric excision of tissue volumes overlying the thermophysically distinct third of the anteromedial tibial diaphysis [16]; and conceptual visualization of high-fidelity 3D thermometric MRI volume generation and TIRI import.](/figures/fig-8-cadaveric-excision-of-tissue-volumes-overlying-the-arrlohqu.png)
Fig. 8. Cadaveric excision of tissue volumes overlying the thermophysically distinct third of the anteromedial tibial diaphysis [16]; and conceptual visualization of high-fidelity 3D thermometric MRI volume generation and TIRI import. ![Fig, 4. SURF interest point descriptor based upon Haar wavelets [11].](/figures/fig-4-surf-interest-point-descriptor-based-upon-haar-33zdu1na.png)
Fig, 4. SURF interest point descriptor based upon Haar wavelets [11]. 
Fig. 5. OpenSURF-based generation of point correspondences and
![Fig. 1. Intertemporal image calibration offset taken from the centroids of the histogram representation of the segmented, emissively uniform, backgrounds of images taken from the same recruit during weeks 1, 4, 7, and 9 of their basic training, as detailed in [5].](/figures/fig-1-intertemporal-image-calibration-offset-taken-from-the-2ml8jzaj.webp)
![Fig. 2. Preprocessed multitemporal TIRI sequence from a symptomatic participant [5].](/figures/fig-2-preprocessed-multitemporal-tiri-sequence-from-a-26kixx96.webp)
![Fig. 8. Cadaveric excision of tissue volumes overlying the thermophysically distinct third of the anteromedial tibial diaphysis [16]; and conceptual visualization of high-fidelity 3D thermometric MRI volume generation and TIRI import.](/figures/fig-8-cadaveric-excision-of-tissue-volumes-overlying-the-arrlohqu.webp)
![Fig, 4. SURF interest point descriptor based upon Haar wavelets [11].](/figures/fig-4-surf-interest-point-descriptor-based-upon-haar-33zdu1na.webp)
Citations
Quantitative Deconvolution of Human Thermal Infrared Emittance
TL;DR: A novel approach to joint inversion of the bioheat transfer model is introduced, exploiting the deterministic temperature dependence of proton resonance frequency (PRF) in low-lipid human soft tissue for characterisation of the relationship between subsurface 3-D tissue temperature profiles and corresponding TIR emittance.
5
References
A threshold selection method from gray level histograms
TL;DR: A nonparametric and unsupervised method ofautomatic threshold selection for picture segmentation is presented, whereby an optimal threshold is selected by the discriminant criterion so as to maximize the separability of the resultant classes in gray levels.
44K
Speeded-Up Robust Features (SURF)
TL;DR: A novel scale- and rotation-invariant detector and descriptor, coined SURF (Speeded-Up Robust Features), which approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster.
14.9K
Evaluation of Features Detectors and Descriptors based on 3D Objects
Pierre Moreels,Pietro Perona +1 more
TL;DR: A method is designed, based on intersecting epipolar constraints, for providing ground truth correspondence automatically, which is based purely on geometric information, and does not rely on the choice of a specific feature appearance descriptor.
An evaluation of open source SURF implementations
David Gossow,Peter Decker,Dietrich Paulus +2 more
- 01 Jan 2011
TL;DR: Different SURF implementations written in C++ are evaluated and it is found that some implementations produce up to 33% lower repeatability and up to 44% lower maximum recall than the original implementation, while the implementation provided with the software Pan-o-matic produced almost identical results.