Professional Documents
Culture Documents
- Database 14
N Treatment
f Website will provide Analysis \
Measurement Tools for: Approaches
12 Knowledge from And Appliances
Textbooks and Scientific 16
Literature
Patient Preferences 24
Photographs
( (Aesthetics, Cost)
Study Models
Server 8 U
Apply Logic-based
rules, Decision
- N
Trees, andfor
Artificial Intelligence 20 Report of Patient
Algorithms Findings, Probabilities
Radiographs to Determine (1) 22 Of Diagnosis and 28
Diagnosis and then Treatment Choices
(2) Treatment Plans, Generated
with both expressed
along with a percent
probability
NT - N -
U.S. Patent Jul. 2, 2013 Sheet 1 of 5 US 8,478,698 B1
U.S. Patent US 8,478,698 B1
8],
U.S. Patent US 8,478,698 B1
9QHORI[15)IH
US 8,478,698 B1
1. 2
METHODS AND SYSTEMIS FOR EMPLOYING BRIEF DESCRIPTION OF THE FIGURES
ARTIFICIAL INTELLIGENCE IN
AUTOMATED ORTHODONTC DAGNOSIS FIG. 1: a diagram illustrating the general steps and pro
AND TREATMENT PLANNING cesses encompassed by the present invention, namely, the
generation of patient data, the analysis of Such data by one or
FIELD OF THE INVENTION more servers, and the automated diagnosis of an orthodontic
condition and the proposed treatment approaches therefor.
The field of the present invention generally relates to meth FIG. 2: a diagram illustrating the various components of
ods and systems that may be used to diagnose an orthodontic the systems described herein, namely, the centralized web
condition. More particularly, the field of the present invention 10 site, server, and database described herein.
relates to methods and systems for automatically diagnosing, FIG.3: a table summarizing the results of a root tip analysis
and proposing a treatment for, an orthodontic condition, of a patient.
which methods and systems employ the use of artificial intel FIG. 4: a table summarizing the results of a tooth torque
ligence capabilities. analysis of a patient.
15 FIG. 5: a table summarizing the results of an arch length
BACKGROUND OF THE INVENTION analysis of a patient.
Many systems and methods have been developed or, more DETAILED DESCRIPTION OF THE INVENTION
typically, envisioned which, hypothetically, could automate
the capture of patient data and diagnosis of an orthodontic The following will describe, in detail, several preferred
condition. These actual (or contemplated) systems employ embodiments of the present invention. These embodiments
certain components and Subsystems that may automate the are provided by way of explanation only, and thus, should not
capture of patient data (such as orthodontic images or scans), unduly restrict the scope of the invention. In fact, those of
the transfer of such data to an orthodontist, and/or even the ordinary skill in the art will appreciate upon reading the
interpretation of Such data (or, more typically, discrete por 25 present specification and viewing the present drawings that
tions of such data). However, the currently-available methods the invention teaches many variations and modifications, and
and systems fail to comprise an ability to make decisions that numerous variations of the invention may be employed,
based on interpreted data, in an automated fashion. In other used, and made without departing from the scope and spirit of
words, the currently-available methods and systems do not the invention.
comprise an effective, accurate, and efficient “artificial intel 30 Referring to FIG. 1, according to certain embodiments of
ligence’ capability, in the automated diagnosis and treatment the present invention, automated diagnosis of an orthodontic
of an orthodontic condition. condition begins with the production of patient-specific data,
The present invention addresses these shortcomings of the which may comprise patient photographs 2, study models 4.
currently-available systems for automated orthodontic diag radiographs 6, and/or combinations thereof. The types of data
nosis and treatment. 35 captured for a particular patient may be the same for all
patients, or may be customized for each patient. The “orth
SUMMARY OF THE INVENTION odontic condition.” referenced herein, may generally com
prise an arrangement of a patient's teeth that is undesirable
According to certain aspects of the present invention, according to applicable orthodontic standards, whereby Such
methods and systems for diagnosing and identifying a treat 40 arrangement may be undesirable for medical, orthodontic,
ment for an orthodontic condition are provided. Such meth aesthetic, and other reasons. Examples of Such orthodontic
ods and systems generally comprise the use of a server on conditions include, but are not limited to, overbites, cross
which a centralized website is hosted. The server is config bites, openbites, overjets, underbites, and the like.
ured to receive patient data through the website, with such These patient data may then be provided to a server 8
patient data comprising patient photographs, study models, 45 through a centralized website 10. Referring to FIG. 2, such
radiographs, and/or combinations thereof. The methods and data may be provided to the server 8 vis-a-vis an on-line form
systems further comprise the use of a database that includes or (within a centralized website 10) through which the data may
has access to (i) information derived from textbooks and be uploaded and transferred to the server 8, or through a
scientific literature and (ii) dynamic results derived from constant data feed through a standard Internet connection. As
ongoing and completed patient treatments. 50 described herein, the server 8 will preferably comprise certain
The invention provides that at least one computer program tools 12 for analysis and interpretation of such data—and for
will operate within the server, which is capable of analyzing making intelligent and probabilistic diagnosis and proposed
the patient data and identifying at least one diagnosis of the treatments for an orthodontic condition.
orthodontic condition (based on the information derived from The invention provides that the server 8 will preferably be
textbooks and scientific literature, dynamic results derived 55 capable of communicating with at least one database 14 (or
from ongoing and completed patient treatments, or combina group of databases). The database 14 will preferably store
tions thereof). The methods and systems further comprise and/or have access to knowledge and information derived
assigning a probability value to at least one diagnosis, with from Scientific, medical, and orthodontic textbooks and lit
the probability value representing a likelihood that the diag erature 16. More particularly, the invention provides that a
nosis is accurate. According to Such embodiments, the meth 60 single database 14 may store all of Such information—or,
ods and systems of the invention further comprise instructing alternatively, it may store portions of Such information and
the computer program to identify at least one treatment the server 8 may have access to additional information that
approach, a corrective appliance, or a combination thereof for may be stored within other databases.
the at least one diagnosis. According to certain preferred embodiments, the invention
The above-mentioned and additional features of the 65 will preferably employ a systematic approach to evaluating
present invention are further illustrated in the Detailed the strength of scientific evidence that may be retrieved from
Description contained herein. the database 14 described herein, for the purpose of diagnos
US 8,478,698 B1
3 4
ing an orthodontic condition (as described below). For The invention further provides that a user may input patient
example, the server 8 may consider the quality, quantity and preferences 24 and/or orthodontist-specified preferences to
consistency of the evidence to derive a grade or confidence the server 8 (through the centralized website 10). For
level of the available knowledge. The invention provides that example, the invention provides that a patient may filter the
various criteria, such as indirect Supporting evidence, may be 5 proposed treatments and corrective appliance results 26 based
taken into account in assessing the strength of each piece of on cost, or the relative aesthetics of an appliance. Similarly, an
scientific evidence. The scientific evidence may then be orthodontist may filter the proposed treatments and corrective
ranked, based on the grade levels (or confidence levels) appliance results 26 based on his/her bias—e.g., an orthodon
assigned thereto. tist may instruct the server 8 to only consider, or to not
More particularly, for example, the invention may consider 10 consider, a certain type of corrective appliance. Upon
the strongest evidence (i.e., evidence of higher grade levels) completion of the foregoing process, the server may be
being derived from at least one systematic review of one or instructed to generate a report 28, which preferably Summa
more well-designed and randomized controlled trials. The rizes the patient data, the diagnoses and associated probabil
invention provides that a second highest grade may be ity values, the proposed treatment approaches and/or correc
assigned to, for example, evidence derived from at least one 15 tive devices (and the probability values associated therewith),
properly designed randomized controlled trial, which and any patient and orthodontist preferences that were con
involved an appropriate sample size and Statistical power. The sidered during the analysis.
invention further provides that a third highest grade may be According to certain embodiments, the invention provides
assigned to evidence derived from well-designed trials, with that the server 8 is configured to analyze the patient data by
out randomization; a single group pre-post, cohort, time identifying a location and position of a plurality ofteethin the
series study; or matched case-controlled studies. Still further, patient data in two-dimensional space or, even more prefer
the invention provides that a fourth grade may be assigned to ably, in three-dimensional space (provided that the type and
evidence from well-designed, non-experimental studies, car amount of patient data provided to the server 8 is sufficient to
ried out by more than one center or research group. A fifth and do so). The invention provides that the server 8 may be con
lowest grade of evidence may consist of opinions of respected 25 figured to undertake this analysis automatically or, according
authorities (which are based on clinical evidence), and/or to certain embodiments, the centralized website 10 will pro
descriptive studies or reports of expert committees. vide users with certain on-line tools to specify the location
The invention provides that the database 14 will further and position of the plurality of teeth in the patient data. For
comprise, or have access to, information that represents example, such on-line tools may be used to identify, within
dynamic results from ongoing and previously completed 30 the patient data, the location and position of a patients inci
orthodontic studies 18. Preferably, these dynamic results 18 sors, canines, premolars and molars, as shown within the
will be organized by orthodontic condition, such that the most patient data that has been provided to the server 8. The loca
relevant information may be retrieved as quickly as possible, tion, position, contours, and size of the plurality of teeth may
within the database 14. Similar to the information derived be mapped out by such user within the centralized website 10,
from Scientific, medical, and orthodontic textbooks and lit 35 while the user is viewing the patient data that has been
erature 16, the invention provides that all of the dynamic uploaded to the server 8, e.g., using a graphics tool that allows
results 18 may be stored within the database 14 or, alterna a user to, for example, approximately trace or identify the
tively, portions thereof may be stored within the database 14 outer boundaries of each tooth.
and other dynamic results 18 may be retrieved, as needed, According to such embodiments, the server 8 may be fur
from third party databases. 40 ther configured to assign coordinates to each tooth within the
Upon providing the server 8 with the patient data, e.g., plurality of teeth. The invention provides that such coordi
patient photographs 2, study models 4, radiographs 6, and/or nates are preferably correlated to the location and position of
combinations thereof, a user may instruct the server 8 to each tooth, as automatically determined by the server (or as
conduct an automated diagnosis. The automated diagnosis otherwise identified by a clinician, using the on-line patient
will be based upon the patient data, the information derived 45 data analysis tools, described above). According to these
from scientific textbooks and literature 16, and dynamic embodiments, the invention provides that the coordinates for
results from ongoing and previously completed orthodontic each of the plurality of teeth may then be compared (by the
studies 18. The server 8 will preferably employ the use of server 8) to a table contained within the database 14. The table
logic-based rules and decision trees 20 to diagnose an orth will preferably comprise a series of diagnostic data sets, with
odontic condition based on all of such information. The 50 each diagnostic data set comprising coordinates, or a range of
invention provides that the server 8 will preferably express coordinates, which are correlated with (1) a known location
the diagnosis by identifying one or more orthodontic condi and position of a plurality of teeth and (2) a previously diag
tions, along with a probability value for each orthodontic nosed orthodontic condition (which previous diagnoses are
condition. According to Such embodiments, the probability derived from (a) information derived from textbooks and
value would represent the relative probability that the diag 55 scientific literature and (b) dynamic results derived from
nosis is accurate. ongoing and completed patient treatments).
Still further, the server 8 will be configured to output (rec According to such embodiments, the server 8 may then be
ommend) one or more treatment approaches and/or corrective instructed to identify a diagnostic data set contained within
orthodontic appliances 22. More particularly, for each diag the database 14 that represents a statistical “best fit,” or most
nosis 20 identified by the server 8, the server 8 will propose 60 closely resembles, the coordinates for the plurality ofteeth of
one or more treatment approaches, corrective appliances, or the patient. At this point, the server 8 may be instructed to
combinations thereof 22. The invention provides that each diagnosis the orthodontic condition based on the “best fit”
Such proposed treatment approach and corrective appliance diagnostic data set that it identified. As mentioned above, the
will be correlated with a probability value. The invention server 8 may further assign a probability value to this diag
provides that this probability value will represent the prob 65 nosis. The probability value will preferably be based, at least
ability of the proposed treatment approach and/or appliance in part, on a confidence level that has been assigned to the
correcting the diagnosed orthodontic condition. diagnostic data set which the server identifies as the statistical
US 8,478,698 B1
5 6
best fit for the coordinates for the plurality of teeth of the Because the systems and methods of the present invention
patient. This confidence level will preferably be influenced by involve certain input provided by users of the invention, the
the grade level that is assigned to the evidence that Supports a systems and methods are dynamic. As such, the invention
connection between the orthodontic condition that is corre provides that algorithms that employ control theory may be
lated with the particular diagnostic data set, as described 5 employed to solve problems in connection with the orthodon
above. tic diagnosis, treatment planning and prognostication steps
According to certain embodiments, the server 8 or, more described herein. Non-limiting examples of such control
particularly, the computer program housed therein, may be theory methods include: Adaptive control, Hierarchical con
instructed to identify at least one treatment approach, a cor trol, Intelligent control, Optimal control, Robust control and
10 Stochastic control.
rective appliance, or a combination thereof for the at least one
diagnosis that is derived from the patient's data. This step EXAMPLES
may be carried by, for example, instructing the server 8 to
calculate a set of target coordinates, which representa desired Example of Optimization Algorithm for Decision Making
and corrected location and position of each tooth in the plu 15 in Diagnosis and Treatment Planning.
rality of teeth of the patient. Based on the target coordinates, An important aspect of multiple optimization is the han
the current location and position coordinates of the patients dling of human preferences, such as the type of cost-and
teeth, and the diagnosed orthodontic position, the server 8 aesthetic-related preferences that a patient or orthodontist
may be instructed to identify at least one treatment approach, may provide to the system described herein. Although selec
a corrective appliance, or a combination thereof, which will tion or prioritizing alternatives from a set of available options
be effective to reorient the plurality of teeth towards the with respect to multiple criteria termed Multi-Criteria Deci
location and position represented by the target coordinates. sion Making (MCDM) is an effective optimization approach,
The server 8 may further be instructed to calculate a prob in practical applications, alternative ratings and criteria
ability value that is correlated with a relative likelihood of the weights can not always be precisely assessed due to unquan
at least one treatment approach, corrective appliance, or a 25 tifiable, incomplete, and/or unobtainable information—or
combination thereof, being effective to reorient the plurality because of a lack of knowledge that may cause subjectiveness
of teeth to a location and position represented by the target and vagueness in decision performance. As such, the inven
coordinates. tion provides that the application of fuzzy set theory to
According to certain preferred embodiments, the invention MCDM models provides an effective solution for dealing
will preferably employ certain additional algorithms in ana 30 with Subjectiveness and vagueness commonly found with
lyzing patient data, diagnosing orthodontic conditions and clinical information. In such embodiments, the invention pro
probability values therefor, and proposing treatment vides that human preferences from both patient and clini
approaches and corrective appliances (and probability values cian-may be assigned “utility values' in which a scaled real
therefor). By way of illustration, as mentioned above, the number is assigned to indicate its relative importance. The
server 8 may be configured to assign greater value (weight) to 35 resulting weighting vector, which evaluates criteria of deci
existing Scientific and medical knowledge, relative to sion making, is then provided in fuZZylinguistic terms such as
dynamic results from ongoing and completed treatments— very poor, poor, fair, good, and very good.
when diagnosing and providing recommended treatment pro Example of Decision Tree Algorithm for Decision Making
tocols for patients. The following will describe certain non in Diagnosis and Treatment Planning.
limiting examples of algorithms, which may be employed in 40 The invention provides that a decision tree method referred
the processes and systems of the present invention. to as "04.5, which allows for input of continuous numerical
The invention provides that artificial intelligence algo data, is preferably employed in the methods and systems
rithms will preferably be employed in order to create an described herein. The invention provides that, under this
artificial neural network, which will enable the server to per approach, a decision tree may be “learned' vis-a-vis splitting
form the orthodontic diagnosis, treatment planning and prog 45 a source set into subsets, based on an attribute value test. The
nostication steps described herein. The algorithms may uti invention provides that this process may be repeated on each
lize statistical estimation, optimization and control theory derived Subset in a recursive manner, which is completed
methodology, or combinations thereof. In the case of statis when the Subset (at a node) has the same value of the target
tical estimation methods, estimators and estimation methods variable, or when splitting no longer adds value to predic
that may be employed include, but are not limited to, the 50 tions. According to this embodiment, decision trees are used
following: maximum likelihood estimators, Bayes estima for relatively simpler functions as decision-tree learners cre
tors, method of moments estimators, Cramer-Rao bound, ate over-complex trees (overfitting), although pruning may,
minimum mean squared error (also known as Bayes least optionally, be performed to minimize this problem. In addi
squared error), maximum a posteriori, minimum variance tion, concepts that are relatively more difficult to learn are not
unbiased estimator, best linear unbiased estimator, unbiased 55 easily expressed by decision trees-and, in Such case, more
estimators, particle filter, Markov chain Monte Carlo, Kal advanced algorithms will be implemented in the systems and
man filter, Ensemble Kalman filter, and Wiener filter. The methods described herein.
statistical optimization techniques that may be utilized Example of Partially Observable Markov Decision Pro
include single-variable optimizations or, more preferably, cesses (POMDPs) and Variants Thereof.
multi-variable optimization techniques. The statistical opti 60 The invention provides that POMDPs are preferably used
mization methods may include, but are not limited to, the in the clinical applications described herein, particularly for
following: Bundle methods, Conjugate gradient method, decisions that are made based on incomplete information.
Ellipsoid method, Frank-Wolfe method, Gradient descent The invention provides that POMDPs are preferably advan
(also known as steepest descent or steepest ascent), Interior tageous insofar as they facilitate the combination of patient
point methods, Line search, Nelder-Mead method, Newton’s 65 data, e.g., patient data derived from examination, photo
method, Quasi-Newton methods, Simplex method and Sub graphs, radiographs and any other diagnostic aids-as well as
gradient method. the current state of knowledge of the cause-and-effect repre
US 8,478,698 B1
7
sentation from these data and measurements. The invention
provides that feature selection may be performed using pat
tern recognition techniques and, furthermore, the treatment Bolton Analysis (Millimeters
decisions with which to restore the patient to a more desirable Anterior (Bolton 6) mm Posterior (Bolton 12) mm
or ideal state are produced. Maxilla 2 mm deficient 2 mm deficient
Patient Example. Mandible
The following example describes the application of the
processes described hereinto a patient in need of orthodontic The invention provides that a series of image analyses may
diagnosis and treatment. The process begins with the patient 10
then be performed, namely, an image analysis of a patients
undergoing cephalometric radiographic analysis. The data frontal and profile planes.
generated by Such analysis are presented in the table below.
The results captured in this Example are summarized in the
tables below.
15
Measurement Patient