Claims:We claim:

1. A method (100) to maximize a success rate (101) of embryo transfer (102) into a uterine cavity (103) of a woman characterized by comprising:
a. training (104) with first set of one or more combinations of images (105) or videos (106) or higher dimensional representations (107) of uteri (108) from a radiological imaging system (109) based on types of uteri (110), a first machine learning based model (111) with a first set of features (112);
b. classifying (113) uteri (108) based on the training (104) using the first machine learning based model (111) with the first set of features (112) into types of uteri (110);
c. training (114) with second set of one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118), a second machine learning based model (119) with features (120) based on the classification (113) of uteri (108), and
d. computing (121) the MIP point (118) in the uterine cavity (103) based on features (114) from the radiological imaging system (109) using the machine learning based model (119) with features (120) based on the classification (113) of uteri (108).

2. A method (122) to maximize the success rate (101) of embryo transfer (102) into a uterine cavity (103) of a woman characterized by comprising:
a. training (114) with a set of one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118), a machine learning based model (123) with features (124) including the classification of uteri as the types of uteri (110), and
b. computing (121) a MIP point (118) of a uterine cavity (103) based on features (124) from the radiological imaging system (109) using the machine learning based model (123).

3. The methods (100) and (122) according to claims 1 and 2, further comprising:
e. assisting (125) in the navigation (126) to reach the MIP point (118) by displaying current position (127) of the tip of the catheter (128) with the radiological imaging system (109) on a display unit (129), and
f. confirming (130) the reaching of the MIP point (118) for embryo transfer (102).

4. The methods (100) and (122) according to claims 1 and 2, wherein first set of one or more combinations of images (105) or videos (106) or higher dimensional representations (107) of uteri (108) from a radiological imaging system (109) are the same as second set of one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109).

5. The methods (100) and (122) according to claims 1 and 2, wherein machine learning models (111), (119) and (123) are obtained using any of the machine learning algorithms (131) including artificial neural networks, classification algorithms such as KNN (K nearest neighbours), multivariable regression, deep learning methods, deep neural networks, convolution neural networks and any combination thereof.

6. The methods (100) and (122) according to claims 1 and 2, wherein the radiological imaging system (109) is any imaging system having the capability to image a uterine cavity (103) and embryo transfer (102) movements such as current position (127) of the tip of the catheter (128) including ultrasound imaging system (132).

7. The methods (100) and (122) according to claims 1 and 2, wherein higher dimensional representations (107) and (117) of uterine cavities (103) from a radiological imaging system (109) includes 3D, 4D, 5D, 6D and 7D representation (133) of uterine cavity (103).

8. The methods (100) and (122) according to claims 1 and 2, wherein MIP point (118) includes a point of intersection of lines (134) that passing through and parallel to cornua (135) of uterus from both the sides ending in the uterine cavity (103), a nearby point (136) learned from the machine learning model (111) and such other points (137) which maximizes the success rate (101) of embryo transfer (102) into a uterine cavity (103).
9. The methods (100) and (122) according to claims 1 and 2, wherein MIP point (118) includes a point of intersection of lines (134) that passing through and parallel to cornua (135) of uterus from both the sides ending in the uterine cavity (103), a nearby point (138) learned from the machine learning models (119) and (123) and such other points (139) which maximizes the success rate (101) of embryo transfer (102) into a uterine cavity (103) based on the classification of the types of uteri (110).

10. The methods (100) and (122) according to claims 1 and 2, wherein features (112), (120) and (124) include anatomical features (140) of uterus and uterine cavity (103) such as size of uterus and uterine cavity (141) including length, breath, height, diameter, volume, size of fundus (142) including length, breath, height, volume, distance between ovary ducts, distance between MIP and tip of the fundus (143), physiological features of women (144) including age, height, weight, BMI and obstetric history such as GPA (Gravida, Para, Abortus) and radiological features (145) such as colors representing the blood flow in the uterine cavity (103), intensity values, region parameters including coordinates of MIP point (146) apart from the success rate (101) and type of the uteri (110) including the positions of uterus.

11. The methods (100) and (122) according to claims 1 and 2, wherein training (104) with one or more combinations of images (105) or videos (106) or higher dimensional representations (107) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118) a machine learning based model (111) with features (112) comprising:
a. collecting (147) features (112) as datasets along with success rate (101) as training datasets (148), and
b. developing (149) the machine learning based model (111) with output variables (150) as coordinates of MIP point (146) using features (112).

12. The methods (100) and (122) according to claims 1 and 2, wherein training (114) with one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118), a machine learning based model (119) or (123) with features (120) or (124) comprising:
a. collecting (147) features (120) or (124) as datasets along with success rate (101) as training datasets (151), and
b. developing (149) the machine learning based model (119) or (124) with output variables (150) as coordinates of MIP point (146) using features (120) or (124).
13. The methods (100) and (122) according to claims 1 and 2, wherein computing (121) a MIP point (118) of a uterine cavity based on features (120) from the radiological imaging system (109) using the machine learning based model (119) comprising:
a. deriving (152) features (120) from the radiological imaging system (109);
b. applying (153) the machine learning based model (119) with features (120), and
c. obtaining (154) the MIP point (118).

14. The methods (100) and (122) according to claims 1 and 2, wherein computing (121) the MIP point (118) of the uterine cavity (103) based on features (124) from the radiological imaging system (109) using the machine learning based model (123) comprising:
a. deriving (152) features (124) from the radiological imaging system (109);
b. applying (153) the machine learning based model (123) with features (124), and
c. obtaining (154) the MIP point (118) based on the classification of types of uteri (110).

15. The methods (100) and (122) according to claims 1 and 2, wherein assisting (125) in the navigation (126) to reach the MIP point (118) comprising:
a. finding (155) the difference between the current position (127) of the tip of the catheter (128) and the MIP point (118);
b. displaying (156) the possible direction (157) including left, right, up, down, above, below and in an angular fashion on the display unit (129) so that a physician can reach the MIP point (118), and
c. pinning (158) the location (159) of the MIP point (118) on the display unit (129).

16. The methods (100) and (122) according to claims 1 and 2, wherein confirming (130) the reaching of MIP point (118) for embryo transfer (102) comprising:
a. checking (160) the difference between the current position (127) of the tip of the catheter (128) and the MIP point (118) to be zero or asymptotically close to zero and
b. indicating (161) on the display unit (129) ready status (162) for embryo transfer (102) at the MIP point (118).

17. The methods (100) and (122) according to claims 1 and 2, wherein the classification of types of uteri (110) includes normal uterus (163), uterine septum (164), bicornuate uterus (165), uterus didephis (166), unicornuate uterus (167) among other classifications (168) including position based classifications such as anteverted, retroverted, anteflexed, retroflexed and such other orientations of uterus.

18. The methods (100) and (122) according to claims 1 and 2, wherein the MIP point (118) in case of Bicornuate uterus (165) is chosen from the set of points (169) including the point near the broader horn for implantation considering the size of both horn and automatically identifying the broader one and the point near horn which is more vascular or having good sub endometrial blood flow by calculating color pixels in each horn and computing percentage of vascularity in each horn.

19. The methods (100) and (122) according to claims 1 and 2, wherein the MIP point (118) in case of uterus didephis (166) is chosen from the set of points (170) including the point near thicker endometrial for implantation considering the size of both endometrium and automatically identify the thicker one and the point near the endometrium having good sub endometrial blood flow by calculating color pixels in each endometrium and computing percentage of vascularity in each endometrium.

20. The methods (100) and (122) according to claims 1 and 2, wherein the MIP point (118) in case of uterine septum (164) or unicornuate uterus (167) or other classification (168) is chosen from the set of points (171) including the point few centimeters below the fundus such as 1 cm below the fundus.
, Description:A method of assistance for identifying and reaching
maximum implantation potential point

PREAMBLE TO THE DESCRIPTION:
[0001] The following specification particularly describes the invention and the manner in which it is to be performed:

DESCRIPTION OF THE INVENTION
TECHNICAL FIELD OF THE INVENTION
[0002] The present invention is related to a method to maximize a success rate of embryo transfer into a uterine cavity of a woman using a radiological system such as ultrasound system in general.
BACKGROUND OF THE INVENTION

[0003] More than 8 million babies born from in vitro fertilization IVF since the world's first in 1978. According to an analysis, in India, the success rate of IVF ranges from 30% to 35%. Globally, the average IVF success rate is around 40% in young women. Over 2.5 million cycles are being performed every year resulting in over 0.5 million deliveries annually. IVF cost per cycle across the globe varies from few thousands of USD to 10K USD. SO, it is essential make success rate high in IVF.
[0004] According to the invention described in US20120016184 titled method of assessing risk of multiple births in infertility treatments, a multiple birth prognostic tool is used to analyze data in order to predict a multiple birth event in a female human patient undergoing an infertility treatment. The MBP prognostic tool may also be used to enhance the accuracy of diagnostic or prognostic tests that predict embryo viability. The MBP prognostic tool of the present invention may be clinic specific or it may be modified to be used in a multi-clinic approach. The key differences include not utilizing the classification of types of uteri, MIP point computation based on that and appropriate machine learning models.
[0005] According to one other invention described in US20200281560 titled ultrasound imaging apparatus, method of controlling the same, and computer program product, an ultrasound imaging apparatus includes: a probe configured to transmit ultrasound waves and detect echo signals; a display; and at least one processor configured to generate an ultrasound image based on the echo signals, wherein the at least one processor is further configured to obtain a three-dimensional (3D) medical image of a uterus, determine a target position for embryo transfer based on the 3D medical image, obtain a real-time ultrasound image of the uterus, identify the target position in the real-time ultrasound image, and control the display to display the real-time ultrasound image and information about the target position. Though many points apparently look similar, the fundamental differences include the features used in machine learning models, not utilizing the classification of types of uteri, MIP point computation based on that and appropriate machine learning models.
[0006] According to yet another invention described in US20140142425 titled systems, methods, apparatuses, and computer-readable media for image guided surgery, methods, systems, devices, and computer-readable media for image guided surgery are presented. The systems herein allow a physician to use multiple instruments for a surgery and simultaneously provide image-guidance data for those instruments. Various embodiments disclosed herein provide information to physicians about procedures they are performing, the devices (such as ablation needles, ultrasound wands or probes, scalpels, cauterizers, etc.) they are using during the procedure, the relative emplacements or poses of these devices, prediction information for those devices, and other information. Some embodiments provide useful information about 3D data sets. Additionally, some embodiments provide for quickly calibratable surgical instruments or attachments for surgical instruments. It is very generic in nature and not confined to embryo transfer as such.
[0007] According to another invention described in US20140142426 titled systems, methods, apparatuses, and computer-readable media for image management in image-guided medical procedures, methods, systems, devices, and computer-readable media for image management in image-guided medical procedures are presented here. Some embodiments herein allow a physician to use multiple instruments for a surgery and simultaneously provide image-guidance data for those instruments. Various embodiments disclosed herein provide information to physicians about procedures they are performing, the devices (such as ablation needles, ultrasound transducers or probes, scalpels, cauterizers, etc.) they are using during the procedure, the relative emplacements or poses of these devices, prediction information for those devices, and other information. Some embodiments provide useful information about 3D data sets and allow the operator to control the presentation of regions of interest. Additionally, some embodiments provide for quick calibration of surgical instruments or attachments for surgical instruments. It is very generic in nature and not confined to embryo transfer as such. Additionally, the fundamental differences include the features used in machine learning models, not utilizing the classification of types of uteri, MIP point computation based on that and appropriate machine learning models.

[0008] Accordingly, there is a need for a method to maximize a success rate of embryo transfer into a uterine cavity of a woman using a radiological system such as ultrasound system in general using machine learning models.

SUMMARY OF THE INVENTION:
[0009] It is an object of the invention to maximize the success rate of embryo transfer into the uterine cavity of the woman using a radiological system such as ultrasound system using the maximal implantation potential (MIP) point.
[0010] One of the methods of the invention includes the following major steps namely, training with a set of one or more combinations of images, videos or higher dimensional representations of uteri, classifying uteri based on the training using a machine learning based model with a set of features into types of uteri and position of uteri, training based on maximal implantation potential (MIP) point, computing the MIP point of the uterine cavity, assisting in the navigation to reach the MIP point and confirming the reaching of the MIP point for embryo transfer.
[0011] Another method of the invention includes similar steps excepting the combined training with a set of one or more combinations of images, videos or higher dimensional representations of uteri combined with the classifying of types of uteri.
[0012] In the embodiments of the invention, the machine learning models involves using artificial intelligence methods including deep learning for training, classification and computing tasks.
[0013] In another embodiment of the invention, types of uteri are considered to maximize the success rate of embryo transfer into the uterine cavity of the woman using a radiological system such as ultrasound system using the maximal implantation potential (MIP) point.
[0014] In yet other embodiments of the invention, the object of the invention is to set different MIP points based on types of uteri and the positions of uteri.

BRIEF DESCRIPTION OF THE DRAWINGS:
[0015] As the figures are only for the illustrating purpose and not to be construed as limiting cases of the invention or implementation of the invention. It should be remembered that for appropriate variants of the embodiments corresponding to different independent claims figures have been provided accordingly.
[0016] Fig.1 describes an overview of one of the embodiments of method of the invention to maximize the success rate of embryo transfer into a uterine cavity of a woman.

[0017] Fig.2 explains another method of the invention depicting an overview to maximize the success rate of embryo transfer into a uterine cavity of a woman including the classification of uteri.
[0018] Fig.3 narrates an extension of the methods to maximize the success rate of embryo transfer into a uterine cavity of a woman as part of assistance to physicians.
[0019] Fig.4 portraits a set of possible machine learning algorithms used in the invention.
[0020] Fig.5. explains the choices of the radiological imaging systems in the invention.
[0021] Fig.6. narrates MIP point (in general) as used in the invention.
[0022] Fig.7 corresponds to possible set of features used in the machine learning models of the invention.
[0023] Fig.8 deals with the details of training step used in the method of the invention.
[0024] Fig.9. corresponds to details of computing step used in the method of the invention.

[0025] Fig.10. corresponds to details of assisting step used in the method of the invention.

[0026] Fig.11. corresponds to details of confirming step used in the method of the invention.

[0027] Fig. 12 explains a sample classification of uteri in the invention.

DETAILED DESCRIPTION OF THE INVENTION:
[0028] Fig.1 describes an overview of one of the embodiments of a method (100) to maximize the success rate (101) of embryo transfer (102) into a uterine cavity (103) of a woman. The major steps include training (104) with first set of one or more combinations of images (105) or videos (106) or higher dimensional representations (107) of uteri (108) from a radiological imaging system (109) based on types of uteri (110) and the positions of uteri, a first machine learning based model (111) with a first set of features (112), classifying (113) uteri (108) based on the training (104) using the first machine learning based model (111) with the first set of features (112) into types of uteri (110), training (114) with second set of one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118), a second machine learning based model (119) with features (120) based on the classification (113) of uteri (108), and computing (121) the MIP point (118) of the uterine cavity (103) based on features (114) from the radiological imaging system (109) using the machine learning based model (119) with features (120) based on the classification (113) of uteri (108). In this method, classification of uteri (108) is done first followed by the estimation of MIP point (118).

[0029] Fig. 2 explains another method (122) to maximize the success rate (101) of embryo transfer (102) into a uterine cavity (103) of a woman including the classification of types of uteri (110). In this method, training (114) with a set of one or more combinations of images (115) or videos (116) or higher dimensional representations (117) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118), a machine learning based model (123) with features (124) including the classification of uteri as the types of uteri (110) and the positions of uteri, and computing (121) a MIP point (118) of the uterine cavity (103) based on features (124) from the radiological imaging system (109) using the machine learning based model (123). In this method, the classified uterus is also provided as one of the parameters. Additionally, based on the classified uterus, different machine learning models can also be generated to estimate MIP point.

[0030] Fig. 3 describes an extension of the methods (100) and (122) to maximize the success rate (101) of embryo transfer (102) into a uterine cavity (103) of a woman as part of assistance to physicians. The two tasks are common to both of the methods. The steps here include assisting (125) in the navigation (126) to reach the MIP point (118) by displaying current position (127) of the tip of the catheter (128) with the radiological imaging system (109) on a display unit (129), and confirming (130) the reaching of the MIP point (118) for embryo transfer (102). The physician performing the embryo transfer (102) will be guided step by step to reach MIP point (118) and confirming (130) the same.

[0031] Fig. 4 deals a set of possible machine learning algorithms (131) used in the machine learning models (111), (119) and (123) of the invention. They include artificial neural networks, classification algorithms such as KNN (K nearest neighbours), multivariable regression, deep learning methods, deep neural networks, convolution neural networks and any combination thereof. It may be recalled that the list of machine learning algorithms mentioned here is not exhaustive and should not be construed that other is not applicable. However, care must be taken to use the appropriate algorithm for classification and regression accordingly.

[0032] Fig.5. describes the choices of the radiological imaging systems (109) in the invention. It deals with any imaging system having the capability to image a uterine cavity (103) and embryo transfer (102) movements such as current position (127) of the tip of the catheter (128) including ultrasound imaging system (132). The higher dimensional representations (107) and (117) of uterine cavities (103) from a radiological imaging system (109) includes 3D, 4D, 5D, 6D and 7D representations (133) of the uterine cavity (103) which are supported by the appropriate ultrasound imaging system (132).

[0033] Fig.6. describes MIP point (118 in general) as used in the invention. The MIP point (118) includes a point of intersection of lines (134) that passing through and parallel to cornua (135) of uterus from both the sides ending in the uterine cavity (103), a nearby point (136) learned from the machine learning model (111) and such other points (137) which maximizes the success rate (101) of embryo transfer (102) into a uterine cavity (103). Additionally, MIP point (118) includes a nearby point (138) learned from the machine learning models (119) and (123) and such other points (139) which maximizes the success rate (101) of embryo transfer (102) into a uterine cavity (103) based on the classification of the types of uteri (110).

[0034] Fig.7 corresponds to possible set of features (112), (120) and (124) used in the machine learning models (111), (119) and (123) of the invention. Appropriate decision must be taken based on the machine learning model which features have to be considered. The features include anatomical features (140) of uterus and uterine cavity (103) such as size of uterus and uterine cavity (141) including length, breath, height, diameter, volume, size of fundus (142) including length, breath, height, volume, distance between ovary ducts, distance between MIP and tip of the fundus (143), physiological features of women (144) including age, height, weight, BMI and obstetric history such as GPA (Gravida, Para, Abortus) and radiological features (145) such as colors representing the blood flow in the uterine cavity (103), intensity values, region parameters including coordinates of MIP point (146) apart from the success rate (101) and type of the uteri (110) including the positions of uterus. As the list is not exhaustive any feature which maximizes the success rate (101) of embryo transfer (102) into a uterine cavity (103) of woman can be always considered as included in the features mentioned here in the invention.

[0035] Fig.8 deals with the details of training step used in the method of the invention. The training (104) refers to training used in the machine learning models (111) with one or more combinations of images (105) or videos (106) or higher dimensional representations (107) of uterine cavities (103) from a radiological imaging system (109) based on maximal implantation potential (MIP) point (118) with features (112). The major steps include collecting (147) features (112) as datasets along with success rate (101) as training datasets (148), and developing (149) the machine learning based model (111) with output variables (150) as coordinates of MIP point (146) using features (112). It is also similar for other trainings used in the invention.

[0036] Fig.9. corresponds to details of computing (121) step used in the method of the invention. The computing (121) refers to computing a MIP point (118) of a uterine cavity based on features (120) from the radiological imaging system (109) using the machine learning based model (119). It consists of the steps namely, deriving (152) features (120) from the radiological imaging system (109), applying (153) the machine learning based model (119) with features (120), and obtaining (154) the MIP point (118). It is also similar in the case of computing in other machine learning models used in the invention.

[0037] Fig.10. corresponds to details of assisting (125) step used in the method of the invention. The assisting (125) refers to the navigation (126) to reach the MIP point (118) for the physician. The steps include finding (155) the difference between the current position (127) of the tip of the catheter (128) and the MIP point (118); displaying (156) the possible direction (157) including left, right, up, down, above, below and in an angular fashion on the display unit (129) so that a physician can reach the MIP point (118), and pinning (158) the location (159) of the MIP point (118) on the display unit (129). It may be recalled that the assistance (125) is provided to the physician to reach the MIP point (118) accurately for embryo transfer (102).

[0038] Fig.11. corresponds to details of confirming (130) step used in the method of the invention. The confirming (130) includes checking (160) the difference between the current position (127) of the tip of the catheter (128) and the MIP point (118) to be zero or asymptotically close to zero and indicating (161) on the display unit (129) ready status (162) for embryo transfer (102) at the MIP point (118).

[0039] Fig. 12 explains a sample classification of types uteri (110) in the invention. They include normal uterus (163), uterine septum (164), bicornuate uterus (165), uterus didephis (166), unicornuate uterus (167) among other classifications (168).

[0040] The MIP point (118) in case of Bicornuate uterus (165) is chosen from the set of points (169) including the point near the broader horn for implantation considering the size of both horn and automatically identifying the broader one and the point near horn which is more vascular or having good sub endometrial flow by calculating color pixel in each horn and computing percentage of vascularity in each horn.
[0041] The MIP point (118) in case of uterus didephis (166) is chosen from the set of points (170) including the point near thicker endometrial for implantation considering the size of both endometrium and automatically identify the thicker one and the point near the endometrium having good sub endometrial flow by calculating color pixel in each endometrium and computing percentage of vascularity in each endometrium.
[0042] The MIP point (118) in case of uterine septum (164) or unicornuate uterus (167) or other classification (168) is chosen from the set of points (171) including the point few centimeters below the fundus such as 1 cm below the fundus.
[0043] For the sake of better understanding of the invention, two example scenarios have been mentioned. It should not be construed that the claimed invention of the patent application works only in these scenarios or restricted to the scenarios mentioned in the examples.