FIELD: medicine.
SUBSTANCE: invention refers to medicine, particularly to oncology, and can be used for differential diagnostics of follicular adenoma and follicular thyroid carcinoma. Thyrocele material is sampled by aspiration fine-needle puncture and/or tumour tissue scrapping; the tumour smears are prepared for the cytological analysis; digital images of the tumour smears are made by the image analysis system comprising the incident light microscope, the digital video camera, the computer and the image processor; the digital images are analysed by double-layer computer neural-net program pre-learned to recognise follicular thyroid carcinoma and adenoma from the master images of the smears histologically diagnosed. Herewith the original digital images of the tumour smears are made at 400-power microscope magnification and resolution 1388×1040 pixels. The original images are downsised to 256×256 pixels within space frequency range 1-128 relative units with using the computer program incorporating the pre-processor. Fourier analysis of the original images enables for automatic convert imaging within space frequency range 96-128 relative units to be used for the further analysis. Each master image of the learning sample is associated with the first layer neuron; in convert image space, the first layer neurons estimate Euclidean distance between each master image of the learning sample and the test image. Herewith derived estimations are assigned with an appropriate positive or negative sign regarding the tumour class (type) of the master image. Minimal Euclidean distance between the master and test image indicate the gainer among the first layer neurons in each of two classes; the only second layer neuron ensures summarising reciprocal Euclidean distances with the related sign within the gain groups. By comparing the total to the limit, the class of the tested image corresponding to follicular adenoma or follicular thyroid carcinoma is specified.
EFFECT: improved diagnostic objectivity and elimination of its dependence on insufficient skills.
1 tbl, 1 ex
Title | Year | Author | Number |
---|---|---|---|
METHOD FOR DIFFERENTIAL DIAGNOSTICS OF FOLLICULAR ADENOMA AND FOLLICULAR THYROID CANCER | 2005 |
|
RU2293524C2 |
METHOD FOR DIFFERENTIAL DIAGNOSTICS OF FOLLICULAR ADENOMA, FOLLICULAR ATYPICAL ADENOMA, FOLLICULAR CANCER AND FOLLICULAR VARIANT OF PAPILLARY THYROID CANCER | 2005 |
|
RU2300319C2 |
METHOD OF DIFFERENTIAL DIAGNOSTICS OF ATYPICAL FOLLICULAR ADENOMA AND FOLLICULAR CANCER IN THYROID GLAND | 2007 |
|
RU2332172C1 |
METHOD OF EARLY AUTOMATED REMOTE DIAGNOSTICS OF SKIN MELANOMA | 2018 |
|
RU2685469C1 |
SYSTEM FOR DIAGNOSING MALIGNANT AND BENIGN SKIN NEW GROWTHS | 2022 |
|
RU2814539C1 |
METHOD OF DIAGNOSING CANCER IN PATIENTS WITH PRE-OPERATION CYTOLOGICAL DIAGNOSIS "FOLLICULAR NEOPLASM" OF THYROID GLAND BY MEANS OF MATHEMATICAL MODELLING | 2011 |
|
RU2493770C2 |
DIFFERENTIAL DIAGNOSTIC TECHNIQUE FOR INDIVIDUAL'S THYROID NEW GROWTHS | 2014 |
|
RU2569154C1 |
DIAGNOSTIC TECHNIQUE FOR DEGREE OF THYROID CARCINOMA | 2011 |
|
RU2485517C2 |
METHOD FOR MICROSCOPIC EXAMINATION OF A SPECIMEN | 2006 |
|
RU2318201C1 |
DIFFERENTIAL DIAGNOSTIC TECHNIQUE FOR THYROPHYMA | 2008 |
|
RU2386398C1 |
Authors
Dates
2009-04-27—Published
2007-05-08—Filed