Please, keep it in mind that there are 3 posts this week. Dont`t wait till the end of week.
Artificial neural network (ANN) are finding many uses nowadays, especially in the field of classification, pattern`s recognition and predicting problems. This powerful tool found medical diagnosis application too. Two medical cases are studied in the paper. The first one is acute nephritis disease, whereas the second one hearth disease. The paper consists of 4 parts including: introduction, ANN general description, experimental results and conclusions.
Scanning the introduction eleven short summaries of ANN medical application, ranging from heart to liver disease, are described. It can be seen that scientist researched application of different ANN models (e.g. supervised, unsupervised, MLP-multilayer perception) and their combination or comparison to a computing methods, technique (e.g. generalized regression, fuzzy approach, genetic algorithms).
The basics of ANN origin and proposed diagnosis model are shown in the second chapter. For example, we are finding out that ANN is a computational model, network of highly interconnecting parallel processing elements (neurons) inspired by human brain. Scientist proposed feed-forward ANN (in this type of network information always moves one direction - never goes backward) for classification, forecasting and solving considered medical problem.
The goal of scientist was to create two neural network model to perform diagnosis of acute nephritis and heart disease. The data set contained 120 samples (symptoms, images, signals), whereas 90 to train the model and 30 for model testing. He evaluated model based on neural network Matlab toolbox. The summary of experimental results is that scientist got very good results - network was able to classify 99% testing cases in acute nephritis disease and 95% in heart disease.
So, let`s start the discussion.
1. Do you regard ANN application in medical diagnosis right?
2. In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
3. Having read undermentioned text give your opinion (3 sentences only).
In the future there will be an on-line application connected with your home input devices. You will be able to take a temperature, do a blood test, input symptoms through the input devices. Then the on-line doctor will make a diagnosis, prescribe medicines or give you a hospital referral.
Dear English seminar contributors,
ReplyDeletePlease, publish your comments. Don't wait till the end of the week as we are to deal with three scientific papers. There are only 3 pages to read.
1 - Do you regard ANN application in medical diagnosis right?
ReplyDelete> It's a very exciting and developing area that using machines to help with medical diagnosis. ANN requires significantly large data to generalize the structures, and patterns. IMHO, the problem is that ANN wouldn't be useful for anything beyond the general issues (weak AI) hence I don't know how we could trust that with our lives. Indeed, there are cases it might work for searching a pattern but nothing to exaggerate.
2. In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
> I believe it's very early for both of them. On the other hand, such an online tool could be handy.
3. Having read undermentioned text give your opinion
> Hopefully we won't loose our privacy at all. If the 'online doctor' term is referring to a virtual actor, it will require the discovery of strong AI. I have a feeling that we need to advance in medical science / biology etc. first since the problems emerging faster than the solutions we develop.
That`s right. Although ANN is a pretty old computing technique (first artificial model of human and animal neuron were proposed by McCulloch & Pitts in 1943) is still very exciting and developing area, especially in medicine.
DeleteI agree up to a point with your opinion in the second part of the first question and the first part in the second one. There are thousands of diseases, so a million of symptoms. I agree that nobody replace doctor. That is out of the question. But, if the ANN perform a support function in medical diagnosis only, don`t replace doctor, nothing bad won`t happen, will it?
I agree with you entirely that a very strong AI is needed to create a virtual doctor and that we need to improve our knowledge in medical science and biology first. But on the other hand if scientists want develop IT in medicine they are not allowed to wait too long, are they?
Q: But, if the ANN perform a support function in medical diagnosis only, don`t replace doctor, nothing bad won`t happen, will it?
DeleteA: I'm not a medical expert but (knowing the limitations and exceptions) I wouldn't like to use such a system for my own and loved ones for now.
Q: But on the other hand if scientists want develop IT in medicine they are not allowed to wait too long, are they?
A: Surely.
Thanks for the article and subject.
Thanks for comments.
DeleteDear seminar contributors,
ReplyDeleteThere are 3 days to left. Only 2 comments were posted. Please don`t wait till the end of the scheduled time. I`m looking forward to your comments.
1. Do you regard ANN application in medical diagnosis right?
ReplyDeleteYes, why not? It's not that much different than taking a survey and getting a score based on symptoms. In either case a physician should look at the results and decide what to do. When IBM's Watson computer won a quiz show against top players, there was speculation that it may be used in a similar application in the medical field. In fact it is already being used. These tools are a great aid, although as Obastemur pointed out above, there may be limits to this particular technology.
2. In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
It depends on the success rate of the system. If it's almost perfect at diagnosing a disease then in theory it should be enough to reliably help a patient. But for now with laws and the experimental stage of these systems it's probably not possible.
3. In my opinion specialized medical equipment at a clinic or a hospital will always be much better than any personal devices. So to get an accurate diagnosis you must have access to that equipment. On the other hand, personal testing may be useful for periodic checkups to detect illnesses sooner.
BTW Medical papers are some of the hardest to read without a medical background (although this one was published under computer science).
Q1: Yes, it`s out of the question than after medical test physician should take a look at results and make decision. I haven`t heard about IMB`s Watson computer before. To the best of my knowledge you dont`t programme Watson computer but you work with it. According to IBM a new generation of doctor are helping Watson learn a language of medicine but in term Warson are helping teach doctors by providing possible treatment [source: http://www.ibm.com/smarterplanet/us/en/ibmwatson/what-is-watson].
DeleteMy quick research of the net shows that Watson bases on cognitive technology. It means that the approach aggregates many of technologies, where ANN are one of many. It means that stand-alone ANN could not do it. Am I right?
Q 2.3. I agree with you entirely that detailed exams required specialized equipment and medical knowledge, as diagnosis is the most important step in the appropriate treatment. So, this approach should be treated only as an introduction or indicator to the next one.
DeleteThank you very much for your comments.
About Watson - that's right, as I've read, Watson works on textual content to answer questions. It's more like a supercomputer because it relies on computing power and lots of data, not just clever algorithms. I guess existing medical descriptions of illnesses could be used as input. Then it would be able to match symptoms to illnesses.
DeleteBut according to Wikipedia they actually used it for something else as the first commercial application, namely: utilization management. So, they used it to optimize the choice of medical treatments versus their cost to get the best healthcare results rather than asking it what the diagnose should be (that is still left to a medical professional).
It is very interesting topic and in my opinion is closely related to Big Data Analysis:)
ReplyDelete1.Do you regard ANN application in medical diagnosis right?
I think that any kind of method that is able to assist doctors with making decisions is good but can not be taken as an indication. I my opinion ANN should be used only as a support for the doctors and as additional information which allows them to make decisions
2.In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
In my opinion ANN should serve only as a support for the doctors because the human body is so complex and sympotomy disease is not predictable enough so we can not delegate decision about diagnosis to some kind of algorithm , doctors may take into account the results of ANN but in the end they should consciously decide.
3.Having read undermentioned text give your opinion (3 sentences only).
I think that this is very good solution and can help and safe many lives.
Imagine situations that you have a heart attack and you are alone and nobody knows about that so nobody will call the ambulance and your chances of survival are minimal but thanks to such a device you can be rescued because it will call an ambulance in time. Of course you need to be aware of what data will be collected such a device in order not to lose privacy
The database of 267 SPECT (Single Proton Emission Computed Tomography) image sets was processed to extract features during heart disease diagnosis. Do you think it is a Big Data Analysis?
DeleteQ1.2.
DeleteI can see, that your opinion coincides to the majority in the way that ANN should only perform a supporting function. I`m solid with your line that because of the complexity of human body some diseases are not predictable by a group of symptoms. That`s way the doctor should explore every avenue based on each source of information which allows them to make decisions and rescue life.
Q3. All in all I thought about standard tests e.g. taking a temperature, doing a blood test, entering symptoms through the input device to the application. Then you could get early diagnosis form "virtual doctor" and then contact with a real first aid doctor.
DeleteI can see, that you went further and imagined a kind of facility connected to the body sensors. The sensors are checking primary body function constantly for example blood pressure, heart rate, etc. Then, if some parameters were wrong and a patient weren`t able to take action the device would send an information to the nearest hospital with the basic information and localisation.
Do I think correctly?
Do you regard ANN application in medical diagnosis right?
ReplyDeleteArtificial neural networks provide a powerful tool to help doctors to analyze, model and make sense of complex clinical data across a broad range of medical applications. Most applications of artificial neural networks to medicine are classification problems; that is, the task is on the basis of the measured features to assign the patient to one of a small set of classes.
In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
we don't must use the machine only for treating patient. Machine made to help the man, not to take it up.
Having read undermentioned text give your opinion.
we can use this technology for diagnostic yes, but the problem here the online doctor.
how we can be sure about this diagnostic its correctly or not.
I think the technology will be the good way for finish the humanity
It is a fact that ANN are mainly directed to classification problems in medical application. Author emphasized it in the introduction too.
DeleteHaving read your opinion to the Q2.3 I can see, that in your view ANN are more as a support in a real medical diagnosis but less than a possible application in a virtual, preliminary medical tests?
1. Do you regard ANN application in medical diagnosis right?
ReplyDeleteI agree with Wiktor and Krzysztof, ANN could assist human doctors in making the right decision about diagnosis of patients, but since they are not that perfect yet they should not decide autonomously. Patients health is too important to be put under such risk, at least not untill the technology/methods mature enough.
2. In your opinion ANN should only support doctors in medical diagnosis or could it work as a stand-alone on-line application?
Just like the answer I provide to the above question, currently ANN should be used only as a supportive tool. However that doesn't mean that it should only be restricted to doctors using it. Such a stand-alone online application could be accessed by patients to at least give them early feedback about non-life-threatening conditions, such as common cold etc. Of course the patient should be wanred beforehand that the results might be false and he should consider a visit to his/her physician. Maybe such a solution could decrease queues to health centers.
3. Having read undermentioned text give your opinion (3 sentences only).
In the future there will be an on-line application connected with your home input devices. You will be able to take a temperature, do a blood test, input symptoms through the input devices. Then the on-line doctor will make a diagnosis, prescribe medicines or give you a hospital referral.
I am sure this will become standard in every household just like internet is nowadays. But for that to happen, we need the tools and technology to mature and become affordable. For now we can only dream and wait. :)
Michail, thank you very much for your line.
DeleteQ1.2. Your point of view correspond to the majority of the opinions. We all agree that ANN should only play a supportive role i medical diagnosis, whereas their functions as a stand-alone application should be limited to non-life-threating conditions, as you underlined.
Q3. We are aware, that nowadays we can only dream on a idea shortly described in Q3 because of the shortage of technology.
DeleteOn the other hand telemedicine is revolutionizing USA healthcare by saving time and money. The idea is focused on non-life-threatening diseases. I agree with you that the tools and technology is needed and it must be checked and mature enough to be popularised. But if the way of entering, transmission and symptoms analysis improve the non-life-threatening diseases could be treated by the "online doctor". Don`t you think it started???
1. Do you regard ANN application in medical diagnosis right?
ReplyDeleteI specialize with neural networks and my answer is absolutely yes! This is for now the most promising method for building AI that can work as an expert system to solve many actual problems. With appropriate power (number of nodes and layers) and training it can think as a human being. Let look at the Blue Brain project.
2. In you opinion ANN should only support doctors in medical diagnosis or could work as a stand-alone on-line application?
We should use this technology for self-diagnosis as we shouldn't use Wikipedia. Of course this technology becomes very accurate but still there is a lot ways for error when this diagnose is made by person without medical background. Not everyone knows that before ANN there was expert system called Mycin which ware design to identify bacteria causing severe infections, such as bacteremia and meningitis, and to recommend antibiotics, with the dosage adjusted for patient's body weight. This system ware developed in 1970s. We still didn't use this system alone without experienced doctor as we do not for example draw our blood for medical examination without nurse but in future current method can become obsolete. However there is one advantage of ANN over the human: It never feels tired so results are always the same for same input.
3. Having read undermentioned text give your opinion (3 sentences only).
This article is only an introduction what ANN can do. ANN can handle huge amount of data and give diagnosis very fast and accurate. If someone doubt what ANN is capable of I recommend an articles about Deep Blue or mentioned earlier Watson.
Piotr, thanku very much for your opinion. Having read you answers I can feel that for you ANN have absolutely application in medical diagnosis. After your answer we know that ANN can handle very fast and accurate with huge amount of data.Except that, could you emphasize shortly another advantages of ANN? Where is the power of ANN? Does the architecture or computing abilities of ANN are better than other?
ReplyDeleteThere's lot to say. In brief we can choose between many types of ANN. From simple Perceptrons which are great in image classification to Hopfield and BAM networks which are good in association memory. The real power of ANN is that they can recognize objects from noised data. Like for example driving car on the same road in sunny or hazy days. Speaking about driving. Have you heard about project ALVINN an autonomous car designed by scientist in Carnegie Mellon University. It use only a few nodes and low-resolution camera but still can drive safely up to 20 miles per hour whitout driver. (https://www.ri.cmu.edu/pub_files/pub3/pomerleau_dean_1991_1/pomerleau_dean_1991_1.pdf).
Delete1. ANN is mathematically proven to be a generalized non-linear modeling method. This makes it as valid in application in medical diagnosis or any other field as other widely used statistical methods widely used statistical methods, from linear regression to GARCH to random forests.
ReplyDelete2. ANNs, if proven to work as well in other tests, should go through the implementation cycle typical for a new technology. Firstly, it should be used in parallel to compare ANN results in real life environment with theoretical results and results of doctors. After that ANNs could be transferred to stand-alone applications with doctors’ supervision. Only after successful results in this “trial period”, ANNs could fully go live as human independent solution for medical diagnosis.
3. I fully agree with this statement. Especially that simple versions of such applications already exist. Check smartphone apps for some American clinics – they use automated questionnaires and provide diagnosis and recovery steps for simple diseases such as flu. This vision is slowly becoming the reality.
Wojciech, lots of thanks for your comments. All your answers, but especially to the third question, confirmed our discusion. There is only one step to change automated questionares into automated devices, that will be able to provide diagnosis and recovery steps for non-life-threarened diseases.
ReplyDelete