Today, the pursuit of accuracy and precision in Radiotherapy has led to the recognition of the problems associated with single-modality diagnostic imaging, driving the development of new methodologies to resolve them. Specifically, this project aims to increase the diagnostic accuracy of MRI by creating a new registration pipeline between histopathological images and MRI scans. This not only enhances precision by leveraging the histopathological image but also provides a solid foundation for the future automated prediction of complete pathological response (pCR, or TRG1) in patients with locally advanced rectal cancer treated with neoadjuvant chemoradiotherapy. This project, in fact, represents the initial phase, involving the preparation of inputs such as the pre-processing of MRI and histopathological images and the insertion of information into a JSON file for recognition. Subsequently, through multimodal registration using affine and deformable transformations, it is possible to visualize the image overlay and the transfer of masks, drawn by pathologists, from the glass slide to the MRI. The development of this semi-automatic pipeline, which includes the pre-processing phase and a GUI-based tool developed in Python, lays the groundwork for the future extraction of radiomic and radio-pathomic features. These features will serve as the input for Machine Learning and Deep Learning programs, to be developed after this project, to define the "ground truth" and predict the response to chemoradiotherapy on the MRI. In the future, this new methodology could lead to a more accurate identification of patients with a complete response who are candidates for an organ-sparing approach, thus avoiding surgical intervention.
A oggi la ricerca di accuratezza e precisione in Radioterapia permettono di riconoscere le problematiche legate alla mono-modalità delle immagini diagnostiche andando a sviluppare nuove metodologie per la loro risoluzione. In particolare in questo progetto si cerca di aumentare l’accuratezza diagnostica della RM andando a creare una nuova pipeline di registrazione tra immagini isto-patologiche e la RM; questo permette non solo l’aumento della precisione data dall’immagine isto-patologica ma anche una solida base per la futura predizione automatizzata del raggiungimento della risposta patologica completa (TRG1) nei pazienti con tumori al retto localmente avanzato trattati con radiochemioterapia neoadiuvante. Questo progetto, infatti, rappresenta la prima fase con la preparazione degli input come il pre-processing delle immagini RM e isto-patologiche e inserimento delle informazioni in un file JSON per il riconoscimento. Successivamente, con la registrazio- ne multimodale e mediante trasformazioni affine e deformabile è possibile visualizzare la sovrapposizione delle immagini e lo spostamento delle maschere, disegnate in Anatomia Patologica, dal vetrino alla RM. Lo sviluppo della pipeline semi-automatica comprendente la fase di pre-processing e un tool con GUI sviluppato in Python rappresentano la base per una futura estrazione delle feature radiomiche e radio-patomiche. Quest’ultime saran- no l’input dei programmi di Machine Learning e Deep Learning che verranno sviluppati successivamente a questo progetto per la definizione del"ground truth" e della risposta al trattamento radiochemioterapico sulla RM. Questa nuova metodologia potrebbe portare in futuro ad una più accurata identificazione dei pazienti con risposta completa da candidare ad un approccio di preservazione d’organo, evitando così l’intervento chirurgico.
Sviluppo pipeline per la co-registrazione multimodale tra immagini di risonanza magnetica e istopatologia come ausilio alla predizione della risposta patologica completa nel tumore del retto
MASSIGNAN, ANNALUCE
2024/2025
Abstract
Today, the pursuit of accuracy and precision in Radiotherapy has led to the recognition of the problems associated with single-modality diagnostic imaging, driving the development of new methodologies to resolve them. Specifically, this project aims to increase the diagnostic accuracy of MRI by creating a new registration pipeline between histopathological images and MRI scans. This not only enhances precision by leveraging the histopathological image but also provides a solid foundation for the future automated prediction of complete pathological response (pCR, or TRG1) in patients with locally advanced rectal cancer treated with neoadjuvant chemoradiotherapy. This project, in fact, represents the initial phase, involving the preparation of inputs such as the pre-processing of MRI and histopathological images and the insertion of information into a JSON file for recognition. Subsequently, through multimodal registration using affine and deformable transformations, it is possible to visualize the image overlay and the transfer of masks, drawn by pathologists, from the glass slide to the MRI. The development of this semi-automatic pipeline, which includes the pre-processing phase and a GUI-based tool developed in Python, lays the groundwork for the future extraction of radiomic and radio-pathomic features. These features will serve as the input for Machine Learning and Deep Learning programs, to be developed after this project, to define the "ground truth" and predict the response to chemoradiotherapy on the MRI. In the future, this new methodology could lead to a more accurate identification of patients with a complete response who are candidates for an organ-sparing approach, thus avoiding surgical intervention.| File | Dimensione | Formato | |
|---|---|---|---|
|
Massignan_Annaluce.pdf
accesso aperto
Dimensione
3.54 MB
Formato
Adobe PDF
|
3.54 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/98459