The project implemented consists of a personal computer application capable of processing photos of blisters currently in circulation in the European Union, in order to extrapolate information about the doses taken during therapy, with the purpose of monitoring the patient and increase his drug adherence. The application installed on a personal computer can recognize the withdrawal of a dose of drug from any blister pack. When the time of taking the medicine occurs, the application manages a series of alarms designed to communicate to the user the correct moment in which it must be taken and, by means of reminders, the correct dose and drug to be taken. By inserting the photo of the blister containing the drug, the user will allow the algorithm to process the image in order to: 1. Check if the drug taken is the correct one 2. Check if the dose of the drug taken is the correct one In the event that the assumption is correct, the file containing the information of the therapy will be updated The project is capable to: 1. remind the patient when to take the dose through an acoustic alarm emitted by the personal computer; 2. remind the patient which drug should be taken; 3. remind the patient of the dose to be taken. Compared to other solutions currently on the market, the project created here shows some features that differentiate it, like: 1. Reduced costs: this project uses the photos that can be taken from any smartphone (now a common device in everyday life) and all the necessary material can be found at low costs (a green or red homogeneous A4 paper used as background). 2. Reusability: this thesis project only require images of blisters without the need to use mechanical or electronic parts for the recognition of the pills 3. Ease of use: the user can use the application by simply taking photographs of the blister packs subject of the therapy. The project was tested by shotting in vivo a series of photographs with different types of blisters. The test allows to evaluate the accuracy of the algorithm that manage the elaboration of the images subjected by the different variability that might affect the photos shots during the usage of the application. The test was divided into three parts: a first part necessary to evaluate the accuracy with which the algorithm recognizes the different types of color of the various blisters, a second part useful to evaluate the accuracy with which the different pills are identified on the blisters and the ability to 7 recognize whether they have been used or not and a third part to evaluate the recognition effectiveness by extracting one pill at a time from the blister. The results obtained in the first test showed that the algorithm can recognize the color of the blister in 100% of cases. The second test for the evaluation of the cavity recognition on the blister packs found accuracy values equal to 99.79% of the cases for the blister packs with cavity in white plastic material, 99.96% for the blister packs with cavity in transparent plastic material and 100% for the blister with aluminum cavity. As regards the recognition of the cavity content, the algorithm has obtained accuracy values of 99.58% for white blisters and 100% for blisters with cavities in transparent plastic material and for blisters with aluminum cavities. The third test found only one error making it necessary to repeat a photograph for white blisters when extracting the seventh pill.
Il progetto realizzato consiste in una applicazione per personal computer in grado di elaborare le foto di blister attualmente in circolazione in Unione Europea, in modo da estrapolare le informazioni riguardo le dosi assunte durante terapia, al fine di monitorare il paziente ed aumentare l’aderenza farmacologica. L’applicazione che viene installata su personal computer, è in grado, tramite analisi delle foto fatte dall’utente, di riconoscere il prelevamento di una dose di farmaco da un qualsiasi blister. Quando sopraggiunge il momento dell’assunzione del medicinale l’applicazione gestisce una serie di allarmi atti a comunicare all’utente il momento corretto in cui deve essere assunto e, per mezzo di promemoria, la dose e il farmaco corretto da assumere. L’utente, inserendo la foto scattata al blister contenente il farmaco permetterà l’elaborazione da parte dell’algoritmo dell’immagine in modo da: 1. controllare se il farmaco assunto è quello corretto 2. controllare se la dose assunta è quella corretta Nel caso in cui l’assunzione risultasse corretta, il file contenente le informazioni della terapia verrà aggiornato. Il progetto quindi è in grado di: 1. ricordare al paziente quando assumere la dose attraverso un allarme acustico emesso dal personal computer; 2. ricordare al paziente quale farmaco va assunto; 3. ricordare al paziente la dose da assumere; Rispetto alle altre soluzioni presenti attualmente sul mercato il progetto qui realizzato mostra alcune caratteristiche che lo differenziano quali: 1. I costi ridotti: questo progetto utilizza le foto che possono essere scattate da un qualsiasi smartphone (dispositivo oramai comune nella vita di tutti i giorni) e tutto il materiale necessario può essere facilmente reperito a basso costo (un cartoncino A4 di colore omogeneo verde o rosso da usare come sfondo alle foto). 2. La riutilizzabilità: Il progetto di tesi necessita soltanto di immagini dei blister senza necessità di utilizzare parti meccaniche e o elettroniche per il riconoscimento delle pillole. 3. Semplicità di utilizzo: L’utente potrà usufruire dell’applicazione semplicemente scattando delle fotografie ai blister oggetto della terapia. Il progetto è stato testato scattando in vivo una serie di fotografie a diverse tipologie di blister. Il test ha permesso di valutare la capacità dell’algoritmo, che si occupa di gestire l’elaborazione delle immagini, di far fronte alle diverse variabilità degli scatti fotografici che si incontreranno durante l’utilizzo della applicazione. Il test è stato suddiviso in tre parti: una prima parte necessaria a valutare l’accuratezza con cui l’algoritmo riconosce le diverse tipologie di colore dei vari blister, una seconda parte utile a valutare l’accuratezza con cui vengono identificate le diverse pillole sul blister e la capacità di riconoscere se 5 sono state utilizzate o meno, e una terza parte per valutare l’efficacia di riconoscimento estraendo dal blister una pillola alla volta. I risultati ottenuti nel primo test hanno dimostrato che l’algoritmo è in grado di riconoscere nel 100% dei casi il colore del blister. Il secondo test per la valutazione del riconoscimento delle cavità sui blister ha riscontrato valori di accuratezza pari a 99.79% dei casi per i blister con cavità in materiale plastico bianco, 99.96% per i blister con cavità in materiale plastico trasparente e del 100% per i blister con cavità in alluminio. Per quanto riguarda il riconoscimento del contenuto della cavità, l’algoritmo ha ottenuto valori di accuratezza del 99,58% per i blister bianchi e del 100% per i blister con cavità in materiale plastico trasparente e per i blister con cavità in alluminio. Il terzo test ha riscontrato un solo errore rendendo necessaria la ripetizione di una fotografia per blister bianchi all’estrazione della settima pillola.
Elaborazione di immagini fotografiche al fine di monitorare l'aderenza al trattamento con farmaci in blister
SAGLIO, UMBERTO
2019/2020
Abstract
The project implemented consists of a personal computer application capable of processing photos of blisters currently in circulation in the European Union, in order to extrapolate information about the doses taken during therapy, with the purpose of monitoring the patient and increase his drug adherence. The application installed on a personal computer can recognize the withdrawal of a dose of drug from any blister pack. When the time of taking the medicine occurs, the application manages a series of alarms designed to communicate to the user the correct moment in which it must be taken and, by means of reminders, the correct dose and drug to be taken. By inserting the photo of the blister containing the drug, the user will allow the algorithm to process the image in order to: 1. Check if the drug taken is the correct one 2. Check if the dose of the drug taken is the correct one In the event that the assumption is correct, the file containing the information of the therapy will be updated The project is capable to: 1. remind the patient when to take the dose through an acoustic alarm emitted by the personal computer; 2. remind the patient which drug should be taken; 3. remind the patient of the dose to be taken. Compared to other solutions currently on the market, the project created here shows some features that differentiate it, like: 1. Reduced costs: this project uses the photos that can be taken from any smartphone (now a common device in everyday life) and all the necessary material can be found at low costs (a green or red homogeneous A4 paper used as background). 2. Reusability: this thesis project only require images of blisters without the need to use mechanical or electronic parts for the recognition of the pills 3. Ease of use: the user can use the application by simply taking photographs of the blister packs subject of the therapy. The project was tested by shotting in vivo a series of photographs with different types of blisters. The test allows to evaluate the accuracy of the algorithm that manage the elaboration of the images subjected by the different variability that might affect the photos shots during the usage of the application. The test was divided into three parts: a first part necessary to evaluate the accuracy with which the algorithm recognizes the different types of color of the various blisters, a second part useful to evaluate the accuracy with which the different pills are identified on the blisters and the ability to 7 recognize whether they have been used or not and a third part to evaluate the recognition effectiveness by extracting one pill at a time from the blister. The results obtained in the first test showed that the algorithm can recognize the color of the blister in 100% of cases. The second test for the evaluation of the cavity recognition on the blister packs found accuracy values equal to 99.79% of the cases for the blister packs with cavity in white plastic material, 99.96% for the blister packs with cavity in transparent plastic material and 100% for the blister with aluminum cavity. As regards the recognition of the cavity content, the algorithm has obtained accuracy values of 99.58% for white blisters and 100% for blisters with cavities in transparent plastic material and for blisters with aluminum cavities. The third test found only one error making it necessary to repeat a photograph for white blisters when extracting the seventh pill.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_saglio_umberto.pdf
non accessibile
Descrizione: Tesi_completa
Dimensione
4.44 MB
Formato
Adobe PDF
|
4.44 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/153758