Fast and Objective Histopathology by multimodal
Transcription
Fast and Objective Histopathology by multimodal
Fast and Objective Histopathology by multimodal tissue auto-fluorescence and Raman Spectroscopy Ioan Notingher School of Physics and Astronomy, University of Nottingham, UK Tissue conserving surgery Frederic Edward Mohs (1936) Fixation, sectioning, staining Mohs micrographic surgery 1-2 hours 30-60 minutes 2500 secondary reoperations per year Frozen-section histopathology 1-2 hours Spectral histopatholog y Intra-operative diagnostics Challenges: 1. Objective diagnosis (not based on subjective interpretation of an image by specialised staff). 2. Diagnosis accuracy similar/higher to the current rates of agreement among histopathologists 3. Fast diagnosis of large tissue samples (3-5 cm), with high spatial resolution (50-100m) 4. Cost-effective diagnosis XYZ Stage Molecular Fingerprints Intensity Sample Raman Shift cm-1 SPECTRA CCD DATABASE (Classification Model) Microscope % match L A S E R Spectrograph CLASS (LABEL) Raman Spectral Imaging (raster scanning) Classification Model Class A Class B Class C Raman spectroscopy for diagnosis of BCC ROC (BCC vs all other classes) DATABASE: 550,000 spectra (from 55 patients) Independent validation (target 95% sensitivity): 220,000 spectra (22 patients) Cross-validation: 100% sensitivity, 93% specificity Can we use Raman spectroscopy to do histopathology for skin sections? BCC RMS Diagnosis 20 20 40 40 Muscle M 60 Fat 60 60 Fat F 80 Unknown 80 80 Unknown U 100 Dermis 100 100 Dermis D 120 Inflammation 120 120 Inflammation In 140 140 160 160 Epidermis E 180 180 160 200 Epidermis Substrate 50 100 150 200 200 50 400 μm Muscle Fat B Muscle 140 BCC BCC 40 180 H&E histopathology 20 Dermis Infl. D. Epid. Substr. Unkn. 100 150 200 200 Substrate 50 S 100 150 200 Can we use Raman spectroscopy to do histopathology for thick resections? Nodular BCC Superficial BCC Infiltrative BCC (Scale bars: 400 μm) Healthy tissue Raman spectroscopy for diagnosis breast tumours (ductal carcinoma) (b) (a) (e) (d) (c) DC NST 20 40 Inflamatory. Stroma 60 Fat 80 1 mm 100 Stroma 120 140 Lobules and Ducts 160 180 Substrate 200 20 40 60 80 100 120 140 160 180 200 Raster scanning BCC Muscle Fat Unknown Dermis Inflammation Raman Model Epidermis Substrate 1mm For 1×1cm2 tissue sample 20 μm spatial resolution 250,000 spectra !!!! (2 s/spectrum => 5.78 days) Multimodal tissue auto-fluorescence and Raman Spectroscopy Looking for plum tomatoes… I cannot find… Ideal technique: -high spatial resolution -- very fast -- does NOT need high specificity for BCC Have you checked the fresh vegetables stand ???! 1600 points Multimodal tissue auto-fluorescence and Raman Spectroscopy 2×2mm2 100 × longer !!! Scale bar: 2mm Kong et al PNAS 2013 110 (38), 15189-15194 Receiver operating characteristic (BCC versus all other classes) Scale bar: 2mm Scale bar: 2mm PNAS 2013, 110 (38), 15189-15194 Diagnosis of BCC for un-sectioned tissue layers Confocal autofluorescence Segmented image MSH diagnosis image BCC 100 Muscle 200 Fat Dermis 300 Inflamed D. 400 Epidermis 500 Substrate Unknown. 600 Scale bar: 2mm 100 200 300 400 500 600 Diagnosis of BCC for un-sectioned tissue layers BCC 500 100 BCC Muscle 1000 Muscle 1500 Fat 200 Fat 2000 Unknown Dermis 300 2500 Dermis 3000 400 Inflammation 500 Substrate Unknown. Epidermis Inflamed D. Epidermis 3500 4000 4500 600 5000 500 1000 1500 2000 2500 PNAS 2013, 110 (38), 15189-15194 3000 3500 100 200 300 400 500 Substrate Conclusions Anaesthetic Excision Tissue sectioning, staining Subjective diagnosis Stop Frozen-section histopathology YES Clear ? NO 45-120 min 5-15 min 1-2 hours Anaesthetic Excision Objective Diagnosis Stop YES Clear ? Spectral histopatholog y NO 2-5 min Acknowledgments Contributors Adrian Ghita Fazliyana Faabar Dr Marta Larraona-Puy Dr Chris Rowlands Dr Kenny Kong Prof Hywel Williams Prof Ian Ellis Dr William Perkins Department of Dermatology Dr Sandeep Varma Dr Iain Leach Dr Emad Rakha Dr Alexey Koloydenko Funding: School of Molecular Medical Sciences, UoN Department of Pathology Queens Medical Centre Nottingham University Hospital NHS Trust Mathematics Department, Royal Holloway University of London