Session Type: 1-hour ePoster Review
Session Title: 1-hour ePoster Review
Authors(s): R. Grande (1), C. Casati (2), D. Di Martino (2), M. Gismondo (3, 1), R. Alberto (1)
Authors Affiliations(s): (1) UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze, Italy, (2) Metasystem Italy, Italy, (3) Università degli Studi di Milano, Italy
Background:
Malaria is one of the major public health challenge across the world, with more of 400.000 deaths for year.Etiological agents of Malaria are five Plasmodium species and the "Golden Standard" diagnosis is the microscopic evaluation of haemoscopy.All but two human Plasmodia are self limited, but the other two, P. falciparum and P. knowlesi, causes severe lethal infections. P. falciparum is the most common imported Malaria in Europe. Sensitivity and specifity of diagnosis by optical microscopy is operator "addicted", thus support of new technologies in this specific field is very interesting for a deep evaluation.This study aim to implement microscopic diagnosis of malaria appling a new technology called AI/deep learning(DL).
Methods:DL algorithm was matched with an microscopic images acquisistion system and a slide scanning system ( Metafer- Metasystem Italy ). It is based on a motorized microscope capable of sampling haematological preparations and digiting an adequate number of optical fields of interest in automation. A batch of 50 P. falciparum positive slides coming from the positive classifier of the UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze( UOC) were used for the deep learning of the system. All the positive slides( thin film and thick drop) were confirmed by ISS( Istituto Superiore di Sanità) experts as current italian Law. All the slides were scanned for stages of P.falciparum and the scanning results were evaluates by UOC experts. The experts corrections were acquired by DL in differents evaluation sessions.
Results:After four working sessions DL showed a 0% false negative microscopic fields rate for P. falciparum and 0.27% false positive microscopic fields for P. falciparum over 1000 scanned microscopic field. The DL system improved ten fold from the first performance.
Conclusions:DL algorithm showed a rising improvement along the learning lessons. The learning performances were very useful and the final results suggest that further studies about capability of the system for identified others Plasmodia species pathogens for humans will be very useful. This diagnostic systems is a promising tool to consolidate and improve malaria diagnosis in labs with poor experience in this specific topic.
Keyword(s): Deep Learning, Diagnosis, MalariaCOI Institutional Grants: Yes
Session Type: 1-hour ePoster Review
Session Title: 1-hour ePoster Review
Authors(s): R. Grande (1), C. Casati (2), D. Di Martino (2), M. Gismondo (3, 1), R. Alberto (1)
Authors Affiliations(s): (1) UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze, Italy, (2) Metasystem Italy, Italy, (3) Università degli Studi di Milano, Italy
Background:
Malaria is one of the major public health challenge across the world, with more of 400.000 deaths for year.Etiological agents of Malaria are five Plasmodium species and the "Golden Standard" diagnosis is the microscopic evaluation of haemoscopy.All but two human Plasmodia are self limited, but the other two, P. falciparum and P. knowlesi, causes severe lethal infections. P. falciparum is the most common imported Malaria in Europe. Sensitivity and specifity of diagnosis by optical microscopy is operator "addicted", thus support of new technologies in this specific field is very interesting for a deep evaluation.This study aim to implement microscopic diagnosis of malaria appling a new technology called AI/deep learning(DL).
Methods:DL algorithm was matched with an microscopic images acquisistion system and a slide scanning system ( Metafer- Metasystem Italy ). It is based on a motorized microscope capable of sampling haematological preparations and digiting an adequate number of optical fields of interest in automation. A batch of 50 P. falciparum positive slides coming from the positive classifier of the UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze( UOC) were used for the deep learning of the system. All the positive slides( thin film and thick drop) were confirmed by ISS( Istituto Superiore di Sanità) experts as current italian Law. All the slides were scanned for stages of P.falciparum and the scanning results were evaluates by UOC experts. The experts corrections were acquired by DL in differents evaluation sessions.
Results:After four working sessions DL showed a 0% false negative microscopic fields rate for P. falciparum and 0.27% false positive microscopic fields for P. falciparum over 1000 scanned microscopic field. The DL system improved ten fold from the first performance.
Conclusions:DL algorithm showed a rising improvement along the learning lessons. The learning performances were very useful and the final results suggest that further studies about capability of the system for identified others Plasmodia species pathogens for humans will be very useful. This diagnostic systems is a promising tool to consolidate and improve malaria diagnosis in labs with poor experience in this specific topic.
Keyword(s): Deep Learning, Diagnosis, MalariaCOI Institutional Grants: Yes
Session Type: 1-hour ePoster Review
Session Title: 1-hour ePoster Review
Authors(s): R. Grande (1), C. Casati (2), D. Di Martino (2), M. Gismondo (3, 1), R. Alberto (1)
Authors Affiliations(s): (1) UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze, Italy, (2) Metasystem Italy, Italy, (3) Università degli Studi di Milano, Italy
Background:
Malaria is one of the major public health challenge across the world, with more of 400.000 deaths for year.Etiological agents of Malaria are five Plasmodium species and the "Golden Standard" diagnosis is the microscopic evaluation of haemoscopy.All but two human Plasmodia are self limited, but the other two, P. falciparum and P. knowlesi, causes severe lethal infections. P. falciparum is the most common imported Malaria in Europe. Sensitivity and specifity of diagnosis by optical microscopy is operator "addicted", thus support of new technologies in this specific field is very interesting for a deep evaluation.This study aim to implement microscopic diagnosis of malaria appling a new technology called AI/deep learning(DL).
Methods:DL algorithm was matched with an microscopic images acquisistion system and a slide scanning system ( Metafer- Metasystem Italy ). It is based on a motorized microscope capable of sampling haematological preparations and digiting an adequate number of optical fields of interest in automation. A batch of 50 P. falciparum positive slides coming from the positive classifier of the UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze( UOC) were used for the deep learning of the system. All the positive slides( thin film and thick drop) were confirmed by ISS( Istituto Superiore di Sanità) experts as current italian Law. All the slides were scanned for stages of P.falciparum and the scanning results were evaluates by UOC experts. The experts corrections were acquired by DL in differents evaluation sessions.
Results:After four working sessions DL showed a 0% false negative microscopic fields rate for P. falciparum and 0.27% false positive microscopic fields for P. falciparum over 1000 scanned microscopic field. The DL system improved ten fold from the first performance.
Conclusions:DL algorithm showed a rising improvement along the learning lessons. The learning performances were very useful and the final results suggest that further studies about capability of the system for identified others Plasmodia species pathogens for humans will be very useful. This diagnostic systems is a promising tool to consolidate and improve malaria diagnosis in labs with poor experience in this specific topic.
Keyword(s): Deep Learning, Diagnosis, MalariaCOI Institutional Grants: Yes
Session Type: 1-hour ePoster Review
Session Title: 1-hour ePoster Review
Authors(s): R. Grande (1), C. Casati (2), D. Di Martino (2), M. Gismondo (3, 1), R. Alberto (1)
Authors Affiliations(s): (1) UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze, Italy, (2) Metasystem Italy, Italy, (3) Università degli Studi di Milano, Italy
Background:
Malaria is one of the major public health challenge across the world, with more of 400.000 deaths for year.Etiological agents of Malaria are five Plasmodium species and the "Golden Standard" diagnosis is the microscopic evaluation of haemoscopy.All but two human Plasmodia are self limited, but the other two, P. falciparum and P. knowlesi, causes severe lethal infections. P. falciparum is the most common imported Malaria in Europe. Sensitivity and specifity of diagnosis by optical microscopy is operator "addicted", thus support of new technologies in this specific field is very interesting for a deep evaluation.This study aim to implement microscopic diagnosis of malaria appling a new technology called AI/deep learning(DL).
Methods:DL algorithm was matched with an microscopic images acquisistion system and a slide scanning system ( Metafer- Metasystem Italy ). It is based on a motorized microscope capable of sampling haematological preparations and digiting an adequate number of optical fields of interest in automation. A batch of 50 P. falciparum positive slides coming from the positive classifier of the UOC Microbiologia Clinica, Virologia e Diagnostica delle Bioemergenze( UOC) were used for the deep learning of the system. All the positive slides( thin film and thick drop) were confirmed by ISS( Istituto Superiore di Sanità) experts as current italian Law. All the slides were scanned for stages of P.falciparum and the scanning results were evaluates by UOC experts. The experts corrections were acquired by DL in differents evaluation sessions.
Results:After four working sessions DL showed a 0% false negative microscopic fields rate for P. falciparum and 0.27% false positive microscopic fields for P. falciparum over 1000 scanned microscopic field. The DL system improved ten fold from the first performance.
Conclusions:DL algorithm showed a rising improvement along the learning lessons. The learning performances were very useful and the final results suggest that further studies about capability of the system for identified others Plasmodia species pathogens for humans will be very useful. This diagnostic systems is a promising tool to consolidate and improve malaria diagnosis in labs with poor experience in this specific topic.
Keyword(s): Deep Learning, Diagnosis, MalariaCOI Institutional Grants: Yes