Table 4 Summary of the impact of image acquisition on the results obtained from dedicated automatic algorithms for image analysis and processing
Object of imaging | Problems in image processing | Measurement problem/error for selected images |
|---|---|---|
Determination of the extent of proliferation in the regenerating rat liver cells using a microscope | Microscope set in different ways – different brightness, focal length | 8% (when measuring the degree of saturation of the colour reaction) |
Thyroid ultrasound images | Thyroid lobe area marked manually in different ways | 2% (difference between experts in the brightness measurement) |
Evaluation of patient’s back temperature distribution using thermographic method | Patients were not told to spread their arms slightly during tests | 31% (average measurement error resulting from grid displacement outside the patient's body) |
Photogrammetric method for assessing postural defects | Effect of incident angle lighting of the patient | 9% one light source, 11% two light sources, 1% four light sources |
Evaluation of the correctness of performing ablative and non-ablative treatments using thermal imaging method in cosmetology | No verification of the places of laser operation (triggered manually) and the dependence of the error on the shape of the skin area subjected to treatment | 18% for the nose, 10% for the cheeks and 7% for the forehead. |
Measurement of the iridocorneal angle in the anterior segment of the eye using a tomograph | Lack of full visibility of the iridocorneal angle | 0-20% depending on the invisible degree of pathology and the amount of the invisible area |
Evaluation of tooth enamel thickness loss using a tomograph | Fragments of the enamel boundary are invisible | 15% (maximum difference in determination of the boundary location between an expert and an automatic method – 50 pixels) |
Tomographic images of the eye fundus - evaluation of layer thickness | Fragments of the retina boundary are invisible | 5% (for such percentage of images not all the layers were fully detected- after the automatic correction) |
Calculation of the eye anterior chamber surface area using a tomograph | Difficulties in correct approximation of the invisible lens with a straight line | 14% (for the location change of the line approximating the lens in the range of ±8 pixels) |
Evaluation of the mechanical properties of the cornea and intraocular pressure in Corvis device | Fragments of the corneal outer contour boundaries are invisible | 47% (due to difficulties in assessing dynamic behaviour of the corneal contour) |