Académique Documents
Professionnel Documents
Culture Documents
A R T I C LE I N FO A B S T R A C T
Keywords: Background and objective: The diagnosis of intestinal parasitosis disease relies on physiological symptoms and
Parasitic diseases diagnosis stool examination. Often, few specialists are available, and manual stool exam is slow, prone to error, and can
Distance regularized level set evolution cause eye fatigue. Our aim was to design and implement a medical expert system that would be automated and
Neuro-fuzzy classifier helpful for diagnosis of human intestinal parasitosis.
Expert system
Methods: The system was developed based on a decision algorithm. A knowledge base was constructed through
information gleaned from books and physicians with information pertaining to the disease. The user interacts
with the system by answering questions. The symptoms information collected led to a microscopic examination
of stools, which was run on the system to detect parasites. The paradigm for automated microscopic examination
of stools consisted of a combined distance regularized level set evolution, automatically initialized by a circular
Hough transform, and a trained neuro-fuzzy classifier. The neuro-fuzzy classifier was trained for analysis of
twenty human intestinal parasites.
Results: We combined the reasoning scheme of diagnosis and the automated clinical exam of stools in the same
system. The parasites found in microscopic imagery confirmed the suspicious disease. The final recommendation
of diagnosis was then completed, with appropriate proposed therapy. The system was evaluated with sixty cases
of infection, and compared to the diagnosis of two expert doctors; we obtained fifty eight correct diagnoses,
corresponding to a 96.6% accuracy.
Conclusions: The proposed system is automated, since the parameters of segmentation, feature extraction and
classification are set to be computationally guided by the type of suspicious parasite. The system is potentially an
important contribution for medical healthcare assistance.
∗
Corresponding author.
E-mail address: daniel.tchiotsop@univ-dschang.org (D. Tchiotsop).
https://doi.org/10.1016/j.imu.2019.100165
Received 28 November 2018; Received in revised form 26 February 2019; Accepted 26 February 2019
Available online 02 March 2019
2352-9148/ © 2019 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
arrhythmias while taking into consideration aspects that are important system for intestinal parasitosis. The authors combined a snake tech-
to the human expert. However, the system was not designed for real- nique with the Hough transform, to extract parasite features in micro-
time operation. Santosh et al. [3] proposed a diagnosis expert system scopic imagery. The probabilistic neural network used for recognition
for medical consult following fast questioning, and offers in addition of the parasites achieved a 100% recognition rate. However, their
the explanation aspects. The system is based on a symbolic inference system was limited to the stage of classification, but may be extended to
engine handling whole rules; thus, it requires a transcription of the diagnostic and treatment proposals. Reviews of another thirteen med-
characteristics and symptoms of the disease to that standard symbolic ical expert systems for diagnosis of various diseases have been carried
system. It returns the problems of the reliability of the data which are out in Ref. [13]. The authors presented for each expert system the di-
set in the system. Soundararajan et al. [4] presented a knowledge based agnosed disease, the technique used, input, as well as remarks.
system for tuberculosis. The system was developed using fuzzy logic for Knowledge Based Systems and Rule Based Expert Systems were recur-
class assignment process and a rule–based fuzzy diagnostics decision rently used. Those systems receive symptoms or results of clinical
support system to assign class labels for tuberculosis. The authors de- exams as input to infer diagnosis. The results this far obtained as de-
veloped 16 rules for conditions and 323 sets of rules for the determi- scribed in the literature suggest that expert systems could be useful to
nation of the class of tuberculosis. The system evaluated the level of risk improve healthcare domains. None of the systems, however, have yet
of tuberculosis patient with received symptoms. However, the system been devoted to human intestinal parasitic diseases.
could not run clinical exams. In this paper, we present an expert system assistant for the diagnosis
Fatumo et al. [5] designed and implemented a medical diagnostic of human intestinal parasite disease. To the best of our knowledge, no
expert system for various kinds of malaria and typhoid complications, expert system has been devoted to the diagnosis of human intestinal
named XpertMalTyph. The system is based on JESS (Java Expert System parasitosis. The system integrates artificial reasoning with an automatic
Shell) programming. Java programming language was used as the im- stools exam. This stools exam is automatic, since parameters of parasite
plementation tool and its Java server page makes the expert system a detection and recognition are set automatically according to suspicious
web-based application. The database engine used was MySQL in- parasites. The system follows a decision algorithm to propose diagnosis
tegrated with JESS. The various modules used have been integrated and appropriate therapy.
from a single web interface. Soltan et al. [6] proposed a medical expert The remainder of this paper is organized as follows: Section II
system for heart disease. The proposed system was constructed in Visual presents the materials and methodology used. In Section III, we present
Prolog 7.3. The symptoms of the patients are obtained by answering some results and discussion. We close with a conclusion.
questions through a user interface. The system then deduced a diagnosis
and proposed a treatment. The system was focused on several heart 2. Materials and methods
diseases, namely angina pectoris and infarction. Mohktar et al. [7] have
developed a rule base expert system to classify the total cholesterol 2.1. Materials
level in the body. Authors used bioelectrical impedance analysis. The
study was investigated on 199 voluntary subjects. The expert system The proposed expert system is implemented on a HP Elite book
was able to correctly classify 6 of 10 subjects, which is a 0.6 probability 6930p with Intel(R) core (TM) 2 Duo CPU, 2.53 GHz, 2 Go RAM, with
accuracy, for classifying total cholesterol level. However, no clinical MATLAB 2014a, on Windows 7. The knowledge base includes rules and
exam could be carried out by the system. The expert system presented facts obtained from relevant books [14,15] and medical doctors. The
in Ref. [8] assists doctors, nurses, and students in orthopedic diseases automated exam of stools includes a pre-process, a segmentation pro-
diagnosis. The system provides a repository of information vis-à-vis cess, and a classification process. We have trained a neuro-fuzzy clas-
multiple diseases. The system was constructed using a tool of expert sifier with twenty classes of recurrent parasites. We have constructed a
systems called a ‘shell’ that allows generating the decision tree, and database of microscopic images gleaned from web-sites [16–18]. We
incorporated expert knowledge with ES_Buider 3.0. The latter generates firstly obtained a database of 1240 original images. In order to increase
a user interface and has an exporting function for web pages creation. the size of our database of images, we have added ‘gaussian’, ‘poisson’,
The authors obtained as results from the expert system information ‘salt & pepper’ and ‘speckle’ noise to each image. We have also rotated
similar to skilled doctors in the domain. Nevertheless, the system was each image with angles of 300, 600, 900, 1200 and 1500. A total of 560
unable to run a test, and only analyzes information from answered new images have then been added to the original database. We have
questions. obtained a comfortable database of 1800 microscopic images corre-
Recently, Nkuma-Udah et al. [9] presented a medical diagnosis sponding to 90 specimens for each of the twenty classes.
expert system to exactly diagnose malaria and delineate its related
diseases in developing countries, using Nigeria as a case study. The 2.2. The expert system
expert system was designed using CLIPS. Doctors and patients tested the
system, and gave a feedback of possible relevant additions to the system The proposed expert system architecture as presented in Fig. 1, in-
application. The system only collected information as symptoms to volves the knowledge base, the decision algorithm, the explanation
produce a diagnosis: no clinical examination can be run with this tool. mechanism and the user interface. It presents the flow of information
A multidisciplinary diagnosis system was presented by Nfongourain exchange between the system and its surroundings: firstly with a
et al. [10]. The system used a probability model for the prediction of the knowledge engineer, who collects rules and facts from doctors and
best individual physicians who are assigned to the case of a patient for books to feed the knowledge base of the system, secondly with the user
diagnosis. They obtained an accuracy of 80% with a multinomial lo- from whom it receives symptoms and proposes a diagnosis.
gistic regression model. Again, the system only addresses information
sharing, and could not analyze or use them for diagnosis. Tchagna et al. 2.2.1. User interface
[11] presented two architectures for a complete processing of biome- The user interface enables communication between the user and the
dical image data using big data analytics. The authors provide in each system. It takes input to the system (symptoms or signs), and presents
step an optimal method that has a big data architecture solution using output to the user, such as diagnosis results, explanation, treatment
Hadoop and Spark frameworks. The system, based on the tools of big details, and recommendations. Symptoms and signs of diseases col-
data technology, analyzes imaging information for biomedical diag- lected leads to recognition of suspicious parasites. Our new system of
nostics. Their system was limited to the stage of classification and the automated examination of stools is based on the detection and re-
analysis of images, which is a restricted to a part of medical diagnosis cognition of intestinal parasites in microscopic images. The method
processes. Saha et al. [12] presented an automated medical diagnosis includes both segmentation and recognition.
2
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
Fig. 2. Extraction process: (a) original image with Entamoeba histolytica trophozoïte, (b) edge map (c) the parasite located, (d) the parasite detected, (e) the
extracted parasite and (f) visualization of histogram oriented gradient features grid locations.
Fig. 3. Extracted microscopic images of parasites. Cyst of: (a) Entamoeba histolytica, (b) Entamoeba coli, (c) Balantidium coli, (d) Chilomastix mesnili, (e) Giardia
lamblia. egg of: (f) Ancylostoma duodenale, (g) Diphyllobothrium latum, (h) Trichuris trichura, (i) Hymenolepis nana, (j) Schistosoma mansoni, (k) Schistosoma
haematobium, (l) Heterophyes heterophyes, (n) Ascaris lumbricoides, (m) Clonorchis sinensis, (o) Fasciola hepatica, (p) Blastocystis hominis, (q) Paragonimus
westermani, (r) Taenia solium. trophozoïte of: (s) Entamoeba histolytica trophozoïte, (t) Iodamoeba butschlii.
3
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
Fig. 4. The decision algorithm flow chart of the proposed expert system, illustrating tests involve in the reasoning scheme.
4
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
5
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
Fig. 6. User interface: (a), (b), (c), (d) and (e) dialog for information collection, (f) suspicious parasites.
and the fact considered. The decision algorithm thus explores all of the 3. Experimental results and discussion
rules contained in the knowledge base, seeking for each one of them to
start its consequence. We have implemented the decision algorithm of Assuming the expert system acts in a manner as a physician for a
the proposed expert system in MATLAB. It follows the flow chart pre- patient, the user will answer a few questions about the patient condi-
sented in Fig. 4, which consists of successive tests, depending upon tions. The system begins with general questions, and is guided by an-
received answers. swers regarding suspicious parasites. Fig. 6((a), (b), (c), (d) and (e))
It is completed with the flow chart of the automated stools exam of present a sample of questions asked to the user. The user answers
Fig. 5 involving the segmentation of the microscopic images of stools of questions by filling the provided space, or clicking on a suitable button.
the patient, and the recognition of the detected parasite. Given a da- Questions successively present are oriented by the decision algorithm
tabase of true facts, all rules that match facts in the database are suc- according to answers and conducted to the suspicious parasite
cessively applied. The conclusion is added to the database. This is re- (Fig. 6(f)).
peated until suspicious parasites are deduced, or a new fact is input, or In Fig. 7(a), we present the dialog interface of the system proposing
others expertise is required. The interrupts paths indicate that other to run a stools exam. If the user clicks on the ‘‘why’’ button, an ex-
symptoms can be taken into account. Suspicious parasites are found in planation of the deduction is given (Fig. 7(b)). The suspicious parasite
microscopic images of stools during automated laboratory tests. Their responsible for health trouble are confirmed or not by the examination
recognition is combined with symptoms to produce a diagnostic. of stools. Fig. 7(c) presents the interface of automatic exam of stools.
After loading microscopic images of stools of the patient using the
button ‘‘load image’‘, the user will successively click on ‘‘CHT-DRLSE’‘,
6
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
Fig. 7. User interface. (a) Dialog proposing to run stools exam (b) recalling information for explanation (c) The automatic exam of stools interface (d) diagnostic
result and a proposed therapy.
7
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165
‘‘HOG-LDA’’ and ‘‘NFC’‘. The system will process the segmentation and and reliable care of patients suffering from intestinal parasitosis.
the recognition of the detected parasite. The proposed scheme of seg-
mentation was applied on our database of microscopic images with Conflicts of interest
various kinds of parasites. Results show how robust and accurate the
method is [19]. The parameters of research are set automatically ac- None declared.
cording to suspicious parasites. The user can also set those parameters
for a particular research, and the system will learn from new conclu- Ethical approval
sions. Those parameters are accessible at the right side of the interface
presented in Fig. 7(c). They are: the estimated radius of research for This article does not contain any studies with human participants or
Hough transforms, the weighted area term Ag (ϕ) , the coefficient of the animals performed by any of the authors.
distance regularization term Rp (ϕ) , time step tstep, the coefficient of the
weighted length term Lg (ϕ) and the maximum number of iterations for Acknowledgements
DRLSE. This opens the field to enlarge the research of new parasites.
The database images were divided in two sets for the training and The authors wish to thank the laboratory assistant TAFFOC Guy
testing phases. The results show a perfect classification of twenty in- Aimé and collaborators of regional hospital of Bafoussam (western
testinal parasites (Protozoa and Helminths) in various stages (cyst, egg Cameroon) for their invaluable support and knowledge sharing. The
and trophozoïte). The twenty classes of parasites were classified with a authors are also grateful to JIPDI M.N (PhD) and PEACHAP Bruno A. for
recognition rate of 100% while testing the trained neuro-fuzzy system improving the overall English level within our manuscript.
[22]. The recognized parasite is combined with other symptoms for the
diagnosis of disease, and a therapy is proposed (Fig. 7(d)). The system Appendix A. Supplementary data
was trained to recognize twenty parasites independently of their stage
of evolution. Thus, the corresponding diseases are also well diagnosed. Supplementary data to this article can be found online at https://
It is worth to mention that the diagnosis and therapy are only a re- doi.org/10.1016/j.imu.2019.100165.
commendation for the user and physician non expert in the domain.
We have simulated the execution of our system on sixty cases of Formatting of funding sources
infection. Questions asked by the proposed system with corresponding
answers of patients were presented to two doctors. Images were picked This research did not receive any specific grant from funding
in a database of 210 microscopic images. The system obtained fifty agencies in the public, commercial, or not-for-profit sectors.
eight correct diagnoses concordant to the same decision of the two
doctors. This corresponds to a correct diagnosis rate of 96.6 0 0 . The References
corresponding therapies for each diagnosis were also approved. In the
two other cases, the two doctors and the proposed diagnosis system [1] Buchanan BG, Shortliffe EH. Rule based expert systems: the MYCIN experiments of
were not in accordance. In one case the two doctors provided different the stanford heuristic programming project. Reading, MA: Addison-Wesley; 1984.
ISBN 978-0-201-10172-0.
diagnoses, while in the other, the expert system proposed a diagnosis
[2] Windyga P, Almeida D, Passariello G, Mora-Ciangherotti FA, Coatrieux JL.
different from the agreed diagnosis of the two doctors. The system still Knowledge-based approach to the management of serious arrhythmia in the CCU,
has problems because of contradictions. For example, when the suspi- medical and biological engineering and computing. May 1991;29(3):254–60.
[3] Santosh KP, Dipti PS, Indrajit M. An expert system for diagnosis of human diseases.
cious parasite is not found in the loaded microscopic image of stools, or
Int J Comput Appl 2010;1(13):0975–8887.
another parasite different from the suspicious one is found, this causes [4] Soundararajan K, Sureshkumar S, Anusuya C. Diagnostics decision support system
contradiction. The knowledge base of the system could however be for tuberculosis using fuzzy logic. Int J Comput Sci Inf Technol Secur (IJCSITS) June
increased with facts and rules for remedy. The trained neuro-fuzzy 2012;2(3). ISSN: 2249-9555.
[5] Fatumo SA, Emmanuel A, Onaolapo JO. Implementation of XpertMalTyph: an ex-
classifier can also be updated with new types of parasite for more re- pert system for medical diagnosis of the complications of malaria and typhoid. J
cognition. Comput Eng (IOSR-JCE) February 2013;8(5):34–40. e-ISSN: 2278-0661, ISSN:
2278-8727.
[6] Soltan RA, Rashad MZ, El-Desouky B. Diagnosis of some diseases in medicine via
4. Conclusion computerized experts system. Int J Comput Sci Inf Technol October
2013;5(5):79–90. https://doi.org/10.5121/ijcsit.2013.5505.
In this paper, a medical expert system assistant is presented with a [7] Mohktar SMM, Ibrahim F, Ismail NA. Expert system for non-invasive classification
of total cholesterol level using bioelectrical impedance. 3rd kuala lumpur interna-
purpose of diagnosis of human intestinal parasitosis. The system in- tional conference on biomedical engineering 2006. p. 63–6. https://doi.org/10.
tegrates artificial reasoning and automatic stools exam. The reasoning 1007/978-3-540-68017-8_17.
part of the system is constructed around a decision algorithm which [8] Zuhra FT, Ahmed AA, Mohsin AT. An implementation of expert system for ortho-
pedic patient diagnosis. quaid-e-awam univ. Res. J. Eng. Sci. Technol. 2016;15(1).
used the constructed knowledge base and user answers to prognosticate [9] Nkuma-Udah KI, Chukwudebe GA. Medical diagnosis expert system for malaria and
on the diagnosis. The parameters of segmentation, features extraction, related diseases for developing countries. IEEE 3rd international conference on
and classification are set automatically, guided by the type of suspicious electro-technology for national development (NIGERCON. 2017. p. 24–9.
[10] Nfongourain MR, Tekinzang TM, Tayou DC. The prediction of good physicians for
parasite sought in the microscopic image. The user is only called to load prospective diagnosis using data mining. Inf Med Unlocked 2018. https://doi.org/
a microscopic image of stools of the patient. This increases the ease of 10.1016/j.imu.2018.07.007.
use of the proposed system. The final recommendation of diagnosis is [11] Tchagna KA, Daniel T, Kengne R, Djoufack TZ, Ngo MAA, René T. An optimal big
data workflow for biomedical image analysis. Inf Med Unlocked 2018;11:68–74.
completed with proposed appropriate therapies. We have evaluated our https://doi.org/10.1016/j.imu.2018.05.001.
system on sixty cases of infection with the diagnosis of two doctors, and [12] Saha TB, Daniel T, Valérie LD, Didier W. Towards an automated medical diagnosis
have obtained fifty eight correct diagnosis corresponding to a 96.6 0 0 system for intestinal parasitosis. Inf Med Unlocked 2018. https://doi.org/10.1016/
j.imu.2018.09.004.
rate. The system is also able to explain its reasoning. This confirms the
[13] Jimmy S, Dinesh G, Abhinav B. Medical expert systems for diagnosis of various
efficiency of the system, which could then be used by a non-expert in diseases. Int J Comput Appl May 2014;93(7):36–43. 0975 –8887.
the task of diagnosis, and as a pedagogic tool for novices. The proposed [14] Levy LE. Handbook of basic techniques for the medical laboratory vol 1. WHO;
expert system is a potentially important contribution in medical diag- 1999. (912pp.) & Vol 2 (1000pp.).
[15] World Health Organization (WHO). Training manual on diagnosis of intestinal
nostic assistance. This system can be improved by increasing the size of parasites, schistosomiasis and intestinal parasites unit division of control of tropical
the database with new types of parasites, and the size of the knowledge diseases vol 48p. World Health Organization Geneva; 2004.
base with new facts and rules. The algorithm could be optimized for [16] Pochet C. (1), «Parasites des aliments» available on line on Internet site: http://
bioimage.free.fr/par_image/parasites_aliments.htm. (Date of access: 08.11.2016,
real-time application. The system will therefore be of support in fast
8
O.T. Nkamgang, et al. Informatics in Medicine Unlocked 15 (2019) 100165